POUR L'OBTENTION DU GRADE DE DOCTEUR ÈS SCIENCES
acceptée sur proposition du jury:
Prof. J. A. Schiffmann, président du juryProf. R. Glardon, Dr E. Boillat, directeurs de thèse
Dr Ph. Thévoz, rapporteurDr V. Romano, rapporteur
Dr A. Jacot, rapporteur
Study and Modelling of the Melt Pool Dynamics during Selective Laser Sintering and Melting
THÈSE NO 6826 (2015)
ÉCOLE POLYTECHNIQUE FÉDÉRALE DE LAUSANNE
PRÉSENTÉE LE 26 NOVEMBRE 2015
À LA FACULTÉ DES SCIENCES ET TECHNIQUES DE L'INGÉNIEURLABORATOIRE DE GESTION ET PROCÉDÉS DE PRODUCTION
PROGRAMME DOCTORAL EN SYSTÈMES DE PRODUCTION ET ROBOTIQUE
Suisse2015
PAR
Tatiana POLIVNIKOVA
To Sveta and Nadia
AcknowledgementsForemost I would like to thank my thesis directors, Rémy Glardon and Eric Boillat, for the
possibility to be a part of LGPP during these four years and to work in its calm and friendly
atmosphere. Thank you for your support, for the encouragement and motivation and for the
patience during long discussions.
I am also grateful to the members of the thesis jury, Valerio Romano, Alain Jacot, Philippe
Thévoz and Jürg Schiffmann, for the interest to my work, for the insightful comments, and
for the fact, that the question session during my private defense turned into an interesting
scientific discussion, which allowed me to look at my work at a new angle.
My sincere thanks to Christoph Van Gestel, Jamasp Jhabvala and Pascal Clausen for the fruitful
collaboration. I am also grateful to Marc Matthey for the technical support at all time.
I would particularly like to thank Ioanna Paniara and Corinne Lebet for their help in solving
administrative problems.
I am very much grateful to my fellow labmates, Maryam Darvish, Christos Tsagkalidis and
Nikola Kalentic for the moral support. Working with you was a pleasure.
This work would not be possible without help of my dear family. I cannot say how much I am
thankful to my mother, Olga Polivnikova, for her love and support. And I will always be grateful
to the members of my family, who unfortunately are no longer here, but who significantly
influenced my personality: my beloved grandfather Valentin Leonov and my godfather Boris
Djubua.
I would also like to thank Sergey Kolosov, a good friend of mine, for the given opportunities.
And I have no words to express my gratitude to my Russian friends, Anna Popkova, Nadia
Sergeeva, Yuri Sergeev, Sveta Korotkova and Ilya Koshkin, who supported me all these years,
despite the distance separating us. Actually, this work I dedicate to friendship.
Lausanne, 23 September 2015 T. P.
i
AbstractSelective Laser Sintering (SLS) and Selective Laser Melting (SLM) are parent layer manufac-
turing processes that allow generating complex 3D parts by consolidating layers of powder
material on top of each other. Consolidation is obtained by processing the selected areas using
the thermal energy supplied by a focused laser beam. In SLS partial fusion of powder particles
takes place, followed by a solidification of the created liquid. SLM is essentially the same pro-
cess as SLS, with the difference that the particles are completely molten under the laser beam.
This development is driven by the need to produce near full dense objects, with mechanical
properties comparable to those of bulk materials and by the desire to avoid lengthy post
processing cycles. Identification of the optimal process conditions (so-called process window)
is a crucial task for industrial application of SLS/SLM processes. Operating parameters of the
process are adjusted in correspondence with optical and thermal properties of the processed
material. Nowadays in SLS/SLM there is a tendency to increase the speed of the fabrication
as a consequence of the available higher laser powers. It leads to increase of laser scanning
speeds. In these circumstances, to rely only on experimental investigations in order to adjust
process and material parameters is time-consuming and ineffective. Simulation tools are
strongly needed for the visualization and analysis of SLS/SLM processes.
In SLM the powder grains under the laser are completely molten and form a liquid domain
called melt pool. Evolution of the melt pool during the process, its interaction with the laser,
the substrate and the surrounding non-molten powder strongly affect the quality of the final
part.
The goal of this work is to study the melt pool dynamics by means of a finite-element simula-
tion software, built specially for SLS/SLM. The numerical model is based on the homogeneous
medium hypothesis. It considers the interaction between the laser and the powder material,
the phase transformations and the evolution of the material properties during the process. We
also study the influence of the phase change on the process efficiency.
The macroscopic model is completed by the sub-models, which allow to study at microscopic
level the processes taking place in the powder bed during its laser heating and melting. Melting
of separate powder particles during laser irradiation is studied by means of the improved
Single Grain Model. The capillary phenomena taking place in the powder bed during SLS/SLM
are also considered. The interconnection of powder grains during their melting is approached
by the mechanism of liquid drops coalescence. According to the obtained results, the depth-
dependent sintering threshold for powder materials is proposed.
iii
Acknowledgements
Key words: Selective Laser Sintering, Selective Laser Melting, powder, 18Ni(300) Maraging
Steel, numerical model, finite element, temperature, enthalpy, sintering threshold, process
efficiency, melt pool, stabilization, effective thermal conductivity, absorptivity, Single Grain
Model, plane wave, capillary, liquid drop, coalescence.
iv
RésuméLe frittage sélectif par laser (SLS) et la fusion sélective par laser (SLM) sont des procédés
de fabrication additifs voisins qui permettent de construire des pièces 3D de géométrie
compliquées en consolidant localement des couches de poudre l’une au-dessus de l’autre.
Les zones à consolider sont mises en fusion grâce à l’énergie thermique fournie par un
faisceau laser finement focalisé. La cohésion du matériau est obtenue après un processus de
resolidification complexe. Dans le procédé SLS la fusion des particules de poudre n’est que
partielle. Le procédé SLM est essentiellement le même à la différence près que les particules
sont complètement fondues. Le développement du procédé SLM à partir du procédé SLS
plus ancien a été motivé par la nécessité de produire des objets entièrement denses, avec
des propriétés mécaniques comparables à celles des pièces obtenues en fonderie, usinage
ou forgeage et sans avoir à passer par de longs cycles de post-traitement. L’identification de
conditions opératoires optimales est une tâche cruciale pour l’application industrielle des
procédés SLS/SLM. Dans la pratique, les paramètres du procédé sont adaptés aux propriétés
optiques et thermiques du matériau utilisé. De nos jours, l’augmentation de la puissance des
lasers disponibles sur le marché a permis une augmentation des vitesses de production des
procédés SLS/SLM. Dans ces conditions, les approches purement expérimentales pour la mise
au point des paramètres procédé sont coûteuses et relativement peu efficaces. Les outils de
simulation sont donc devenus nécessaires pour l’optimisation et l’analyse des technologies
SLS/SLM.
Dans le procédé SLM, les grains de poudre sous le laser sont fondus complètement. Ils forment
un domaine liquide appelé bain de fusion. L’évolution de ce domaine au cours du processus,
de son interaction avec le laser, le substrat et la poudre non fondue affectent fortement la
qualité de la pièce finale.
Le but de ce travail est d’étudier la dynamique du bain de fusion au moyen d’un logiciel
de simulation par éléments finis, adapté spécialement aux procédés SLS / SLM. Le modèle
numérique est basé sur la technique mathématique d’homogénéisation du milieu traité et
prend en compte de nombreux mécanismes comme l’interaction entre le laser et la poudre, les
transformations de phase et l’évolution des propriétés du matériau en cours de consolidation.
Nous étudions également l’influence du changement de phase sur l’efficacité du procédé.
Le modèle macroscopique est complété par des sous-modèles, qui permettent d’étudier
des processus qui se déroulent, à l’échelle microscopique, dans le lit de poudre lors de son
traitement par laser. La fusion des particules de poudre en cours d’irradiation est étudiée au
moyen d’une généralisation du modèle classique du grain isolé. Les phénomènes de capillarité
v
Acknowledgements
se produisant dans le lit de poudre pendant les procédés SLS/SLM sont également étudiés.
En particulier, l’interconnection des grains de poudre en cours de fusion est analysée grâce à
des modèles (théoriques et numériques) de coalescence de gouttes. Les résultats obtenus ont
notamment conduit à la définition d’un seuil de frittage dépendant de l’intensité du faisceau
laser et de ses propriétés de pénétration dans la poudre.
Mots clefs : frittage sélectif par laser, fusion sélective par laser, poudre, acier maraging 18Ni
(300), modèle numérique, éléments finis, température, enthalpie, seuil de frittage, efficacité
des processus, bain de fusion, stabilisation, conductivité thermique effective, absorption effec-
tive, modèle du grain isolé, onde plane, écoulement capillaire, goutte de liquide, coalescence.
vi
ContentsAcknowledgements i
Abstract (English/Français) iii
List of figures xi
List of tables xiii
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv
List of symbols xv
1 Introduction 1
1.1 Research objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Thesis structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2 State of the Art 5
2.1 SLS/SLM process definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.1 Binding mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.2 Materials and applications . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.3 Requirements to the process . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Characterization of SLS/SLM process . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2.1 Process parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2.2 Material parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.3 Numerical modelling of SLS/SLM . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3 18Ni(300) Maraging Steel Powder 13
3.1 Grain size distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.2 Bulk material properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.3 Powder properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.3.1 Surface absorptivity and optical penetration depth . . . . . . . . . . . . . 16
3.3.2 Powder density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.3.3 Effective thermal conductivity . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.4 Molten material properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.5 Operation parameters used for Maraging Steel Powder . . . . . . . . . . . . . . . 23
3.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
vii
Contents
4 Single Grain Model 25
4.1 Single Grain Model: main assumptions . . . . . . . . . . . . . . . . . . . . . . . . 26
4.1.1 Intensity of the plane wave and laser power . . . . . . . . . . . . . . . . . 27
4.1.2 Heating of a spherical grain with a plane wave . . . . . . . . . . . . . . . . 28
4.2 Melting of a powder grain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.3 Evolution of powder bed properties . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.4 Different powder grain sizes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5 Melting Dynamics 39
5.1 Capillary properties of molten powder material . . . . . . . . . . . . . . . . . . . 40
5.1.1 Dynamic viscosity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.1.2 Surface tension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.1.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
5.2 Liquid drops coalescence model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
6 Numerical Modelling of SLS/SLM 51
6.1 Numerical model description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
6.1.1 Enthalpy and temperature evolution . . . . . . . . . . . . . . . . . . . . . 52
6.1.2 Sintering potential and effective properties . . . . . . . . . . . . . . . . . 53
6.1.3 Sintering potential evolution . . . . . . . . . . . . . . . . . . . . . . . . . . 54
6.2 Numerical method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
6.2.1 Time discretization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
6.2.2 Space discretization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
7 Sintering Threshold in SLS/SLM 61
7.1 Evolution of the grain melting threshold . . . . . . . . . . . . . . . . . . . . . . . 62
7.2 Process parameters and material properties . . . . . . . . . . . . . . . . . . . . . 65
7.3 Simulation results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
7.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
8 Mechanisms Affecting the Efficiency of SLS/SLM Process 69
8.1 Thermal conductivity and absorptivity evolution in SLS/SLM . . . . . . . . . . . 70
8.1.1 Simulation: constant material properties . . . . . . . . . . . . . . . . . . . 70
8.1.2 Rosenthal solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
8.1.3 Simulation: evolving material properties . . . . . . . . . . . . . . . . . . . 73
8.1.4 Influence of absorptivity evolution on SLS/SLM efficiency . . . . . . . . . 74
8.2 Convection and radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
8.2.1 Convection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
8.2.2 Radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
8.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
viii
Contents
9 Melt Pool Dynamics 83
9.1 Bulk material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
9.2 Powder bed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
9.3 Powder layer on a substrate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
9.4 Theoretical model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
9.5 Stabilization distance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
9.6 Different scanning strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
9.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
10 Conclusions 97
Bibliography 105
Curriculum Vitae 107
ix
List of Figures2.1 The SLS process scheme [1]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.1 18Ni(300) Maraging Steel Powder. . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.2 Grain size distribution of the 18Ni(300) Maraging Steel powder [2]. . . . . . . . . 14
3.3 β-function for 18Ni(300) Maraging Steel. . . . . . . . . . . . . . . . . . . . . . . . 17
3.4 Radiation reflection, transmission and absorption. . . . . . . . . . . . . . . . . . 18
3.5 Discrete thermal resistances model for powder beds by A.V. Gusarov (pic. [3]). . 20
4.1 Irradiation of the surface of a powder bed by a moving laser. . . . . . . . . . . . 26
4.2 Meshing of a 2D domain for the Single Grain Model calculations. . . . . . . . . 28
4.3 Temperature distribution inside a spherical grain after 6 µs of irradiation by the
plane wave with the intensity corresponding to the 100 W laser (Tab. 4.1). . . . 29
4.4 Evolution of molten material proportion during grain melting by a plane wave
of different intensities approaching laser beam with power 100 W, 200 W, 300 W. 30
4.5 Maximum irradiation of a powder grain by a top-hat laser beam. . . . . . . . . . 31
4.6 Evolution of the proportion of the molten material during irradiation of two
neighbouring grains: (i) t < t 0; (ii) t ' t 0, (iii) t > t 0. . . . . . . . . . . . . . . . . . 32
4.7 Temperature evolution in the contact point between two grains irradiated by a
plane wave, approaching the laser of different powers. . . . . . . . . . . . . . . . 33
4.8 Evolution of the average temperature of a powder grain (Tab. 4.1), irradiated by
a plane wave, approaching the laser of different powers. . . . . . . . . . . . . . . 34
4.9 Single Grain Model: Time to the contact spot melting t 0 and average temperature
T 0av of the powder grain at the moment t 0. . . . . . . . . . . . . . . . . . . . . . . 36
5.1 The dynamic viscosity of pure liquid iron in the range of temperatures 1538C <T < 2500C [4]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.2 Coalescence of two liquid drops. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.3 Relative contact spot radius evolution: simulation and theoretical results, r0 =4.435 µm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.4 Relative contact spot radius evolution for different radii of liquid drops: solid
lines – LFEM, dashed lines – Shikhmurzaev’s eq. (5.9) and Thorrodsen’s eq. (5.7). 46
5.5 Contact spot radius evolution between two liquid drops of different radii: (i)
r1 = 2.5 µm, r2 = 4.435 µm; (ii) r1 = 4.435 µm, r2 = 10 µm (eq. (5.11)). . . . . . . 47
xi
List of Figures
6.1 Typical domainΩ: powder bed (with top surface Σ) on a substrate. . . . . . . . 52
6.2 Illustration of the net of parallelepipedic cells for the space discretization of the
sintering potentialΦ. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
6.3 Top view of the non conformal mesh for the temperature/enthalpy profiles. . . 60
7.1 Evolution of the molten material fraction and the average temperature of a
18Ni(300) Maraging Steel powder grain for different laser powers. . . . . . . . . 62
7.2 The dependency of the grain melting threshold T SGMt on the wave intensity. . . 63
7.3 The dependency of the melting threshold on the depth for wave intensities,
approaching laser powers 100 W, 200 W, 300 W. . . . . . . . . . . . . . . . . . . . 64
7.4 Exact laser path and explanation of the laser delay. . . . . . . . . . . . . . . . . . 65
7.5 Time above the sintering threshold when Tt = T SGMt (solid lines) and when
Tt = T f (dashed lines) for Sets I and II (Tab. 7.2). . . . . . . . . . . . . . . . . . . . 67
7.6 EBSD analysis of the results for sets I and II (Tab. 7.2). Provided by Dr. J.Jhabvala. 68
8.1 Influence of the thermal conductivity and absorptivity evolution on the maxi-
mum temperature reached on the surface of the material during laser scanning
(see Tab. 8.1). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
8.2 Absorption of laser power in the beginning of the scanning for 18Ni(300) Marag-
ing Steel powder. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
8.3 Approximate position of the stabilized melt pool with respect to the laser spot:
P = 300 W, υ= 1 m/s, ω= 40 µm (by Mathematica software). . . . . . . . . . . . 78
9.1 Evolution of the melt pool surface area during laser scanning of 18Ni(300) Marag-
ing Steel powder and bulk material (Tab. 3.6). . . . . . . . . . . . . . . . . . . . . 85
9.2 Melt pool evolution during melting of 32µm powder layer on a substrate: P = 100
W, υ= 1 m/s. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
9.3 Temperature evolution according to (9.1): 0.1, 0.3, 0.5, 0.7, 0.9 ms (from the right
to the left) from the laser irradiation start: P = 100 W, υ= 1 m/s. . . . . . . . . . 89
9.4 Scanning strategy used for the numerical modelling in § 9.6. . . . . . . . . . . . 93
9.5 Melt pool evolution during scanning of the strategy Fig. 9.4: P = 100 W, υ= 1 m/s. 94
9.6 Overlapping of scanned tracks for the strategy Fig. 9.4 and top-hat laser beam of
radius 40 µm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
xii
List of Tables3.1 18Ni(300) Maraging Steel chemical composition [2]. . . . . . . . . . . . . . . . . 15
3.2 18Ni(300) Maraging Steel bulk material properties. . . . . . . . . . . . . . . . . . 15
3.3 Thermal conductivity of powders in air at normal conditions: theoretical predic-
tion and experimental results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.4 Process parameters used in the study. . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.5 Sets of the process parameters used in simulation. . . . . . . . . . . . . . . . . . 23
3.6 Properties of 18Ni(300) Maraging Steel. . . . . . . . . . . . . . . . . . . . . . . . . 24
4.1 Properties of the laser and of a typical 18Ni(300) Maraging Steel powder grain. . 27
4.2 Single Grain Model. (i) t s : start of the grain melting (Tab. 4.1); (ii) t l : complete
grain melting; (iii) t imax: maximum time of grain irradiation (Fig. 4.5). . . . . . . 30
4.3 Melting time t 0 of the contact point between two grains, irradiated by a plane
wave of intensity I0, and the average temperature T 0av of the grains at this moment. 33
5.1 Properties of liquid 18Ni(300) Maraging Steel, used in Chapter 5. . . . . . . . . . 42
5.2 Parameter K2 (see eq.(5.7)) for different temperatures and radii of the drops [5]. 44
5.3 Coalescence time of equal liquid drops. . . . . . . . . . . . . . . . . . . . . . . . . 46
5.4 Coalescence time of liquid drops with different radii. . . . . . . . . . . . . . . . . 48
7.1 Melting threshold T SGMt of a 18Ni(300) Maraging Steel powder grain, irradiated
by a plane wave of intensity I0, approaching the laser of power P . . . . . . . . . 62
7.2 Sets of process parameters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
8.1 Influence of the thermal conductivity and absorptivity evolution on the maxi-
mum temperature reached under the laser: simulation vs. Rosenthal solution
(8.4). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
8.2 Stabilization time of the surface state within the laser spot. . . . . . . . . . . . . 75
9.1 Time to stabilize the melt pool for different process and material parameters. . 84
9.2 Melt pool depth reached during laser melting of 18Ni(300) Maraging Steel powder. 86
9.3 Material properties used for the estimation of the melt pool stabilization time. 91
9.4 Melt pool stabilization distance for different process and material parameters. 92
9.5 Simulation parameters for the scanning strategy presented in Fig. 9.4. . . . . . . 93
xiii
List of Symbols
Amp Surface area of the stabilized melt pool
Cp Heat capacity
E dsurf Surface laser energy density
E dvol Volume laser energy density
Ea Activation energy
I0 Incident radiation intensity
Ir Reflected radiation intensity
Lc Characteristic length
L f Latent heat of fusion
Lv Latent heat of vaporization
Mth Powder grain thermal mass
N Mean coordination number
P Laser power
Q Intensity distribution of the heat source
R Reflectivity of the material
Rg Universal gas constant
T Temperature
T0 Temperature of the environment
T 0av Average temperature of the grain at the moment t0 (SGM)
T f Melting temperature
xv
Tt Melting threshold temperature
T SGMt Melting threashold of the powder grain
Tv Boiling temperature
Ω Working domain
Φ Sintering potential
Σp Contact thermal resistance between two powder particles
α Surface absorptivity of the material
αl Surface absorptivity of liquid material
αp Surface absorptivity of the powder
αs Surface absorptivity of solid material
δ Optical penetration depth of the material
δ0 Radius of the sphere, used for temperature averaging
η Thermal diffusivity
γ Liquid-gas surface tension
γa Adiabatic exponent of the gas phase
κ Thermal exchange coefficient
E0 Electric amplitude of the plane wave
st Poynting vector
Bo Bond number
Kn Knudsen number
Nul Nusselt number
Ral Rayleigh number
µ Dynamic viscosity of the liquid
µp Vacuum magnetic permeability
ω Laser beam radius
ρ Density
ρl Liquid material density
xvi
ρp Powder density
ρs Solid material density
σ Stefan-Boltzmann constant
τM Maxwell relaxation time of the liquid
τsint Characteristic sintering time of the material
υ Laser scanning speed
ε Material emissivity
εI Imaginary part of the dielectric constant
εR Real part of the dielectric constant
εd Dielectric constant
ζ Sintering rate
d0 Powder grain diameter
d bmp Melt pool stabilization distance for the bulk material
d plmp Melt pool stabilization distance for the powder layer
d thmp Theoretical stabilization distance of the melt pool
fs Fraction of the solid material
g Gravitational acceleration
hpbmp Melt pool depth in the powder bed
hh Hatching distance
hl Powder layer thickness
k Effective thermal conductivity
katm Thermal conductivity of the surrounding atmosphere
kg Thermal conductivity of gas phase
kl Effective thermal conductivity of liquid material
ks Effective thermal conductivity of solid material
l0 Gas mean-free path
q Heat source induced by Joule heating in the powder grain
xvii
qabs Heat flux corresponding to absorbed heat
qconv Heat flux corresponding to convective heat losses
qrad Heat flux corresponding to radiative heat losses
qsurf Surface heat source
r0 Powder grain radius
rc Contact radius
t Time
t 0 Time of contact point melting (SGM)
t tr Transition time of the surface state within the laser spot
t imax Maximum time of the grain irradiation (SGM)
t l Time of complete grain melting (SGM)
t s Time to melting of the powder grain (SGM)
tatt Time when the melt pool touches the substrate
t bmp Melt pool stabilization time for bulk material
t pbmp Melt pool stabilization time for powder bed
t plmp Melt pool stabilization time for powder layer on the substrate
t thmp Theoretical melt pool stabilization time
tc Time of liquid drops coalescence
u Enthalpy per unit mass
xviii
1 Introduction
Selective Laser Sintering (SLS) and Selective Laser Melting (SLM) are parent layer manufac-
turing processes that allow generating complex 3D parts by consolidating layers of metallic,
ceramic, or polymer powder material on top of each other. Consolidation is obtained by
processing the selected areas using the thermal energy supplied by a focused laser beam. In
SLS partial fusion of powder particles takes place, followed by a solidification of the created
liquid. SLM is essentially the same process as SLS, with the difference that the particles are
completely molten under the laser beam and form a liquid domain called melt pool. The main
advantages of SLS/SLM are three-dimensional geometrical freedom, complying with modern
design requirements, mass customization and material flexibility.
Industrial application of SLS/SLM technologies requires from the process rather strict condi-
tions to be fulfilled: porosity, hardness, surface roughness, geometrical accuracy and other
properties of the final part should be sufficient to guarantee its reliability over the entire life
cycle. The economical aspects – production time and cost – also should be considered.
SLS/SLM operating parameters (so-called process window) are adjusted in correspondence
with optical and thermal properties of the material. Nowadays in SLS/SLM there is a tendency
to increase the speed of the fabrication as a consequence of the available higher laser powers.
Identification of the optimal process parameters is a crucial task for industrial application of
SLS/SLM.
Efficiency of SLM process and the quality of the final part are strongly connected to the
evolution of the melt pool during the process, its interaction with the laser, the substrate
and the surrounding non-molten powder. Knowledge of the melt pool geometry allows to
adjust such important operating parameters as hatching distance of the scanning strategy
and powder layer thickness. Prediction of the melt pool surface area evolution and melt pool
position with respect to the laser helps to estimate and increase process efficiency and can be
also useful for the calibration of online monitoring systems for Selective Laser Melting.
1
Chapter 1. Introduction
Parameters of the melt pool and the character of its interaction with the laser radiation during
SLM are defined by the evolution of thermal and optical properties of the powder due to its
melting. Physical laws, describing this evolution, are related to the phenomena taking place in
the powder at the level of powder grains: to their melting dynamics and their interaction in
the presence of the liquid phase.
We can see that SLS/SLM processes involve dynamical phenomena, strongly dependent
on numerous parameters. To rely only on experimental investigations in this case is time-
consuming, ineffective and expensive. Numerical modelling of SLS/SLM is strongly needed.
All numerical approaches, developed for SLS/SLM, describe a working domain (a powder on a
substrate) with some thermal and mechanical properties and certain boundary conditions
and a heat source of a specified profile shape, which interacts with the domain. One of
the commonly used modelling methods is based on the description of the heat transfer
in the homogenized working domain, characterized by its effective properties. However,
as mentioned above, it is clear that for better understanding of SLS/SLM it is necessary to
consider the process at microscopic level, i.e. at the level of powder grains. Generally it
leads to the increase of numerical model complexity and, consequently, to the decrease of
computation speeds.
The solution to this problem is in the use of a numerical model based on material homoge-
nization, completed with separate sub-models simulating the process at the powder grain
level. The sub-models can be divided in two categories:
• models, describing the effective properties of a loose powder;
• models, studying powder grain melting dynamics and their interaction in the presence
of the liquid phase.
Results of these sub-models, added to the basic macroscopic model, will allow to simulate the
SLS/SLM even in the absence of large computing powers.
2
1.1. Research objectives
1.1 Research objectives
This study will focus on the following objectives:
1. Development of theoretical models for the description of the loose powder effective prop-
erties.
2. Development of theoretical models describing the melting dynamics and the interactions
of molten powder particles during SLS/SLM.
3. Theoretical investigation and numerical simulation of SLS/SLM using a macroscopic
model based on the homogeneous medium hypothesis, completed by the results of micro-
scopic sub-models.
4. Estimation of the influence of material properties evolution on SLS/SLM process efficiency.
5. Study of the properties and dynamics of the melt pool formed during SLM process.
This work is mainly concerned by the study of SLS and SLM of 18Ni(300) Maraging Steel pow-
der. The considered operating parameters are chosen in accordance with modern industrial
requirements.
3
Chapter 1. Introduction
1.2 Thesis structure
The structure of this work is as follows. Chapter 2 introduces main principles of Selective Laser
Sintering and Selective Laser Melting. It also discusses process requirements and describes
commonly used approaches for the numerical simulation of SLS/SLM.
In Chapter 3, 18Ni(300) Maraging Steel is studied. Properties of the solid and liquid phases of
the material are discussed. Different methods for estimating the effective powder parameters
are also introduced. The process parameters, chosen in the context of this study, are also
presented.
In Chapters 4 and 5, we introduce the models studying SLS/SLM at the powder grain level. In
Chapter 4, using the improved Single Grain Model, we study melting of a separate powder
particle. We approach laser irradiation by a plane wave. Melting time and threshold tempera-
ture of a powder grain are estimated. Chapter 5 focuses on the capillary phenomena taking
place in the powder during melting. The dynamics of molten powder grain interconnection is
considered in the framework of the liquid drop coalescence model.
Chapter 6 describes the numerical methods used to discretize the macroscopic SLS/SLM
model. Theoretical assumptions, based on the results of Chapters 4 and 5, are presented.
Recommendations about the space and time discretization are given.
Chapter 7 introduces a model for a depth-dependent sintering threshold, based on the results
of the Single Grain Model (Chapter 4). An experimental validation of this model is also
presented.
Chapter 8 focuses on phenomena, taking place during SLS/SLM, which can influence the
process efficiency. The evolution of powder effective parameters due to the phase change is
studied, convective and radiative losses are estimated.
In Chapter 9 the melt pool dynamics during SLM is discussed. Evolution of the melt pool
during the process is studied and melt pool stabilization criteria are presented. A method
to adjust the powder layer thickness, according to material and process parameters, is also
introduced.
4
2 State of the Art
Selective Laser Sintering (SLS) and Selective Laser Melting (SLM) are parent Additive Manufac-
turing processes using a laser to melt and solidify metallic, ceramic, polymer or composite
powders. Additive manufacturing is the name given to a host of related technologies used to
create physical objects by sequential deposition of material layers. The originality of Additive
Manufacturing is that material is added to form the desired part and not subtracted from a
bigger part. All the non-used material can be recycled in the next construction.
5
Chapter 2. State of the Art
Figure 2.1 – The SLS process scheme [1].
2.1 SLS/SLM process definition
Selective Laser Sintering (SLS) is a layer manufacturing process that allows generating complex
3D parts by consolidating layers of powder material on top of each other [6, 7]. The object
is built layer by layer. Consolidation is obtained by processing the selected areas using the
thermal energy supplied by a focused laser beam (Fig. 2.1). Using a beam deflection system,
each layer is scanned according to its corresponding cross-section as calculated from the CAD
model. Partial fusion of the particles takes place, followed by a solidification of the created
liquid.
Selective Laser Melting (SLM) is essentially the same process as SLS, with the difference that
the particles are completely molten under the laser beam and form a liquid bath (a melt
pool). This development was driven by the need to produce near full dense objects, with
mechanical properties comparable to those of bulk materials and by the desire to avoid lengthy
post-processing cycles.
The main advantages of SLS/SLM are three-dimensional geometrical freedom, complying
with modern design requirements, mass customization and material flexibility. In contrast to
material removal techniques, complex parts can be fabricated without the need for lengthy
tool path calculations. Shapes that may not be realizable by conventional machining can
be built (conformal cooling channels in injection molds) [1]. SLS/SLM technology makes it
possible to create fully functional parts directly from metals, ceramics and plastics without
using any intermediate binders or any additional processing steps after the laser sintering
operation [8, 9, 10, 11]. Remaining unprocessed powder can be reused.
6
2.1. SLS/SLM process definition
2.1.1 Binding mechanisms
Different binding mechanisms can be responsible for the consolidation of the powder during
SLS and SLM processes [12, 13, 14]:
• Solid State Sintering (SSS). Occurs below the material melting temperature: diffusion
of atoms in solid state (volume diffusion, grain boundary diffusion or surface diffusion
[15, 16]) creates necks between adjacent powder particles growing with time [17, 18].
The mechanism is rarely applied as diffusion of atoms in solid state is slow and does not
meet the requirements of the process.
• Liquid Phase Sintering (LPS). Part of the powder material is molten while other parts
remain solid. The liquefied material spreads between the solid particles almost instan-
taneously as it is driven by intense capillary forces. The material that melts may be
different from the one remaining solid: the ‘low melting point’ material is called the
binder, while the ‘high melting point’ material is generally called the structural material.
The binder may remain as part of the final product or may be removed during a debind-
ing cycle. Binder and structural materials can be brought together simply by using a
mixture of two-component powders, or they can be combined in composite or coated
powders.
• Partial Melting. The heat supplied to a powder particle is sufficient only to melt the
grain surface, while particle core remains solid. The molten material plays the role of
the binder. Small necks are formed between the non-molten particle cores [1]. This
binding mechanism is used for metals. The technique is sometimes called Direct Metal
Laser Sintering (DMLS) [15].
• Full Melting. Applied to achieve fully dense parts without need for any post-processing.
One of the most popular technologies nowadays. However, to achieve satisfactory
results, requires a detailed preliminary analysis of the material and process parameters
(so-called process window). For each new material, a process-window needs to be
determined separately, in order to avoid scan track instabilities (sphereodisation of the
liquid melt pool, also known as ‘balling’) and part porosity [19, 20].
2.1.2 Materials and applications
SLS/SLM technologies are widely used in various industries, medicine and research, offering
a range of advantages compared to conventional manufacturing techniques. Most of the
materials, used in SLS/SLM, are developed by machine makers and are strictly application-
oriented [14, 21].
7
Chapter 2. State of the Art
• Metals. Iron-based powders and various steels (moulds and tools [22, 23, 24]), titanium-
based powders and Co-Cr (implants and scaffolds).
• Polymers. Polyamide, polystyrene and PEEK (SLS plastics), PCL (biomedical applica-
tions), PVA, PC.
• Ceramics. HA, SiC, ZrO2, Al2O3, glass-ceramic, Ta2O5, bismuth titanate. Least-processed
materials by SLS/SLM, as far as they are susceptible to develop cracks.
2.1.3 Requirements to the process
Industrial application of Selective Laser Sintering/Melting requires from the process rather
strict conditions to be fulfilled: porosity, hardness, surface roughness, geometrical accuracy
and other properties should be sufficient to guarantee the reliability of the final product over
the entire life cycle. It is also necessary to consider the economical aspects of the process –
production time and cost.
Laser power and laser scanning speed are adjusted in correspondence with optical and thermal
properties of the process material. In the case of polymers and composites the technique is
now well understood and widely used, but for metals and ceramics it still suffers from a lack of
precision, surface roughness and poor mechanical properties. Identification of the optimal
process parameters (process window) is a crucial task for industrial application of SLS/SLM
processes.
According to the industrial requirements, nowadays in SLS/SLM there is a tendency to increase
the speed of the fabrication as a consequence of the available higher laser powers (up to 1 kW).
For metal powders it leads to increase of scanning speeds (up to 3−4 m/s).
8
2.2. Characterization of SLS/SLM process
2.2 Characterization of SLS/SLM process
Material and process parameters, listed in §§ 2.1.2 – 2.1.3, are strongly related to each other.
Laser and scanning parameters are adjusted according to chemical, optical and heat transfer
properties of the processed material.
2.2.1 Process parameters
The relation between laser and strategy parameters defines the amount of energy received
by a processed material and, subsequently, is of a great influence on the product properties.
It would be convenient to have a variable which can link all the process parameters and,
therefore characterize SLS/SLM process.
Laser energy density is a commonly used factor to characterize SLS/SLM process and proper-
ties of a final sintered part. Generally energy density can connect laser power, scanning speed,
hatching distance of the scanning strategy, laser beam diameter and thickness of the powder
layer.
According to [25, 26, 27, 28] surface laser energy density E dsurf (in J/m2) can be expressed as:
E dsurf =
P
υhh, or E d
surf =P
2υω, (2.1)
where P is the laser power, υ – the scanning speed, hh – the hatching distance and ω – the
radius of the laser beam.
An expansion of the expression (2.1) is proposed in [29]:
E dsurf =
P
υhhx, (2.2)
where ratio x = 2ω/hh characterizes the overlapping of the trace of a laser beam spot.
Volume based laser energy density E dvol (in J/m3) is also widely used for the characterization of
SLS/SLM processes [30]:
E dvol =
P
υhhhl, (2.3)
where hl is the thickness of a powder layer.
However, it can be seen, that volumic density of laser energy doesn’t take into account the
diameter of the laser beam and the beam profile. Therefore, it can be reasonable to add
another factor, characterizing the laser – laser intensity or laser fluence.
9
Chapter 2. State of the Art
In fact, no one of the parameters (2.1) – (2.3) reflects all the details of the Selective Laser
Sintering/Melting process due to its non-linearities. We will discuss this problem in detail in
the next chapters.
2.2.2 Material parameters
One of the commonly used methods to characterize a powder is to consider it as a homo-
geneous medium with some effective properties. We will discuss this model in detail in
Chapter 6.
In the framework of the homogeneous medium hypothesis, a powder bed, like any homoge-
neous material, can be characterized by its thermal diffusivity η= k/ρ/Cp , where Cp is the
heat capacity of the material, ρ is the powder bed density and k is the effective value of its
thermal conductivity. Optical parameters of a powder bed are described by its effective surface
absorptivity and optical penetration depth.
During SLS/SLM powder properties evolve to the parameters of a bulk material. In this study
we will discuss the consequences of this evolution.
However it is clear, that the modification of parameters during laser sintering/melting is the
result of changes in the structure of the powder bed. For better understanding of SLS/SLM at
macroscopic level, we need to consider processes, which take place at the level of a powder
particle, like melting, rearrangement and interaction of powder grains. Therefore powder
properties (packing, grain size distribution, etc.) strongly influence the process.
2.3 Numerical modelling of SLS/SLM
The complexity of SLS/SLM processes is that they involve dynamical phenomena, caused by
a time dependent temperature distribution [31]. These phenomena are strongly dependent
on many parameters. Process variables for SLS/SLM technologies can be divided in several
groups:
(a) Material properties: Chemical composition, size distribution, shape, optical and heat
transfer properties of powder and molten material.
(b) Laser parameters: Laser power, spot size, beam spatial distribution, scanning speed,
possibly pulse rate and duration.
(c) Strategy of manufacturing: Scanning strategy, powder deposition parameters, powder
layer thickness, application of a protective atmosphere, preheating.
To rely only on experimental investigations in order to adjust process and material parameters
is time-consuming, ineffective and expensive. Simulation tools are strongly needed for the
visualization and analysis of SLS/SLM processes.
10
2.3. Numerical modelling of SLS/SLM
All numerical approaches, developed for SLS/SLM modelling, describe a powder bed with
some thermo-mechanical properties and certain boundary conditions and a heat source
of a specified profile shape, which interacts with the powder bed. First attempts aiming to
describe theoretically the mechanisms of the heat transfer in a powder bed were made in
[17, 32, 33]. Mathematical models of the heat distribution for single scan tracks were proposed
in [34, 35]. In [36] an analytical model was used to evaluate the temperature field created
in a titanium sample during pulsed laser treatment. Finite element analysis by means of
commercial software was performed in [37] for laser enamelling. A finite element numerical
model, based on the relations between specific enthalpy, temperature and sintering potential,
was introduced in [38] (see Chapter 6 for details). In [39] the transient temperature field for
TiAl6V4 SLM parts was predicted. A 3D model for the analysis of the relations between the
process parameters and the mass of a powder layer was proposed in [40].
All these models, described above, are based on the homogenization of a powder bed. However
the question of interaction of laser radiation with a powder, considered as a packing of particles,
was also studied. Numerical model, based on the solution of Radiation Transfer Equation
(RTE) for the absorbing and scattering medium was proposed in [41, 42]. For the same purpose
ray-tracing modeling was used in [7, 43].
Evolution of effective thermal parameters of a powder bed during SLS/SLM was studied in
[38, 44, 45, 46].
Nowadays there is a tendency to model SLS/SLM processes not only at macroscopic, but also
at microscopic level, aiming to study powder melting dynamics. These models are often based
on the solution of Navier-Stokes equations (for example [47]). However the lattice Boltzmann
method (LBM) becomes more popular [48].
11
3 18Ni(300) Maraging Steel Powder
This work is mainly concerned by the study of Selective Laser Sintering and Selective Laser
Melting of 18Ni(300) Maraging Steel powder. In this Chapter we describe the properties of the
bulk material. We also discuss the effective values of the thermal conductivity and of the light
absorptivity in the loose powder. We introduce different methods to estimate the effective
parameters. For the estimation of the effective powder absorptivity we use a theoretical
approach based on the solution of the radiation transfer equation. For the calculation of the
effective thermal conductivity of a powder bed we apply a discrete thermal resistances model.
We also introduce an experimental validation of these theoretical methods.
Besides that we introduce and discuss process parameters chosen for this study.
13
Chapter 3. 18Ni(300) Maraging Steel Powder
Figure 3.1 – 18Ni(300) Maraging Steel Powder.
3.1 Grain size distribution
The photo of 18Ni(300) Maraging Steel powder we study is shown in Fig. 3.1. The grain size
distribution of the powder can be seen in Fig. 3.2 [2]. In this work the details of the real distri-
bution will not really be taken into account. Instead we will often approach it by a monosized
distribution for a diameter d0= 8.87 µm, which corresponds to a radius r0= 4.435 µm.
Grain size, µm
%n
um
ber
ofg
rain
s
0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0 50.00.0
2.0
4.0
6.0
8.0
10.0
12.0
14.0
16.0
18.0
20.0
d0
Figure 3.2 – Grain size distribution of the 18Ni(300) Maraging Steel powder [2].
14
3.2. Bulk material properties
Aluminum, Al 0.10 %Boron, B ≤ 0.0030 %Carbon, C ≤ 0.030 %Cobalt, Co 9.0 %Iron, Fe 67.0 %Manganese, Mn ≤ 0.10 %Molybdenum, Mo 4.80 %Nickel, Ni 18.5 %Phosphorous, P ≤ 0.010 %Silicon, Si ≤ 0.10 %Sulfur, S ≤ 0.010 %Titanium, Ti 0.60 %Zirconium, Zr 0.010 %
Table 3.1 – 18Ni(300) Maraging Steel chemical composition [2].
Density of the solid material ρs 0.008 g/mm3
Thermal conductivity ks 26 W/m/C
Heat capacity Cp (for T> 1091C) 0.5 J/g/C
Melting temperature T f 1413 C
Boiling temperature Tv 2862 C
Latent heat of fusion L f 247 J/g
Latent heat of vaporization Lv 6260 J/g
Table 3.2 – 18Ni(300) Maraging Steel bulk material properties.
3.2 Bulk material properties
The chemical composition of the 18Ni(300) Maraging Steel bulk material is given in Tab. 3.1.
Other physical properties are presented in Tab. 3.2. According to [2], we assume T f = 1413C
as the material melting point. Due to lack of complete information, they do not all correspond
exactly to 18Ni(300) Maraging Steel. Sometimes they have been interpolated from the proper-
ties of pure iron or close grades of steel. The values L f and Lv correspond to the latent heat of
fusion and latent heat of vaporization of Grade 316 Stainless Steel, respectively [49, 50]. For Tv
we take the vaporization point of iron.
According to [2], the density ρs of 18Ni(300) Maraging Steel is not dependent on temperature.
The thermal conductivity of the material at 50 C is ks = 25.8 W/m/C [2]. We round this
value to 26 W/m/C and assume it to be constant. The heat capacity Cp strongly depends
on temperature. The value given in Tab. 3.2 (Cp = 0.5 J/g/C) will only be used for tempera-
tures T > 1091C. For smaller values of T , the heat capacity can be deduced from the data
concerning the β-function of the material (see Fig. 3.3).
15
Chapter 3. 18Ni(300) Maraging Steel Powder
β-function
In Chapter 6, describing the numerical model, we will use the so-called β-function, which
connects the enthalpy per unit mass u to the temperature T . The β-function of 18Ni(300)
Maraging Steel has been measured by EMPA [51] for u < 598.9 J/g and under the assumption
that β(0) = 0 which means that
u = 0 J/g, when T = 0C. (3.1)
For values of u larger than 598.9 J/g, the β-function is constructed under the assumption
that Cp ' 0.5 J/g/C and by taking the values of L f , Lv , T f and Tv into account (see Tab. 3.2).
The final result is presented in Fig. 3.3. Fig. 3.3b is the zooming of the grey area on the graph
Fig. 3.3a.
3.3 Powder properties
As it will be explained in detail in Chapter 6, we will consider the powder bed as a homoge-
neous medium, which is characterized by effective properties: surface absorptivity, optical
penetration depth, thermal conductivity and density. These parameters strongly depend on
powder geometry and material properties.
In this section we present methods to estimate the effective properties. We also discuss
assumptions we have to make for the 18Ni(300) Maraging Steel powder.
3.3.1 Surface absorptivity and optical penetration depth
The ability of a powder to absorb laser energy is one of the key factors for the SLS/SLM process
quality. In general, when electromagnetic radiation is incident to the surface of a medium, it is
partially reflected from the surface and partially penetrates into the material (Fig. 3.4). The
reflected part of the radiation Ir , if non-linear phenomena are neglected, can be expressed as:
Ir = RI0, (3.2)
where I0 is the intensity of the incident radiation and R is the reflectivity (or the reflection
coefficient) of the material. Further in this study, to characterize the ability of the material to
absorb radiation, we will use the notion of the surface absorptivity:
α= 1−R. (3.3)
In fact, surface absorptivity of a medium is defined as the fraction of the incident radiation
which penetrates into the medium.
16
3.3. Powder properties
Enthalpy u, J/g
Tem
per
atu
reT
,C
0.0 1000 2000 3000 4000 5000 6000 7000 8000 90000.0
700
1400
2100
2800
3500
4200
Tv transition liquid → gasgas
ph
ase
liqu
idp
has
e
Lv
(a) β-function for 18Ni(300) Maraging Steel.
Enthalpy u, J/g
Tem
per
atu
reT
,C
0.0 100 200 300 400 500 600 700 800 900 1000 1100 1200 13000.0
200
400
600
800
1000
1200
1400
1600
1800
2000
trans.solid → liq.T f
liquid
ph
asesoli
dp
has
e
L f
(b) β-function for 18Ni(300) Maraging Steel (zoom of the grey region).
Figure 3.3 – β-function for 18Ni(300) Maraging Steel.
17
Chapter 3. 18Ni(300) Maraging Steel Powder
Figure 3.4 – Radiation reflection, transmission and absorption.
During propagation in the medium, the non-reflected radiation is partially absorbed by
the material. In the case of a plane electromagnetic wave, propagating in a homogeneous
medium along direction z, perpendicular to the medium surface, the radiation intensity decay
according to Beer-Lambert law. The intensity of the transmitted radiation at a distance z from
the material layer surface (see Fig. 3.4) is:
Itr(z) =αI0e−z/δ, (3.4)
where δ is the optical penetration depth of the material, I0 – the incident intensity and α – the
surface absorptivity (see (3.2)–(3.3))
If the incident radiation is not a plane wave but a Gaussian beam (TEM00) the law (3.4)
generalizes under some conditions [52]. These conditions are:
(i) the material surface corresponds the waist plane of the beam;
(ii) the distance z is relatively small compared to the beam radius ω:
z ¿ kλω2, (3.5)
where kλ = 2π/λ is the wave number, λ – the laser wavelength.
In such a situation, one may write:
Itr(x, y, z) 'αI0(x, y)e−z/δ (3.6)
with I0(x, y), the incident gaussian beam intensity at location (x, y) on the top surface (Fig. 3.4).
Equation (3.6) is known as the paraxial approximation of the gaussian beam.
18
3.3. Powder properties
In the case we study λ = 1070 nm, ω = 40 µm (see Tab. 3.4), and the condition (3.5) for the
paraxial approximation is: z ¿ 9.3 mm. Therefore for powder layers up to several hundreds
microns thick the propagation of the gaussian beam can be described by (3.6).
The optical penetration depth is a measure of light propagation in the material. It is defined as
the depth at which the radiation intensity inside the material falls to 1/e of its original value at
the surface. Generally, for bulk metals, the optical penetration depth is of the order of tens of
nanometers [53]. As we can see from (3.4), if the optical penetration depth is small compare to
the thickness of the material δ¿∆z, we can neglect the transmitted part of the radiation Itr.
The incident radiation falling onto a powder bed is not only partially absorbed by the outer
surface but also penetrates through the pores due to multiple reflections. For metallic powders
with grain size of 10− 20 µm, δ is of the order of several powder particle diameters [54].
Therefore in laser sintering/melting of a powder layer with a thickness less than hundred
microns, the penetration depth plays an important role in laser energy deposition. The
influence of δ increases for highly reflective metallic powders [43].
Estimation of the parameters of laser energy deposition onto a powder bed was an objective of
many studies [7, 17, 41, 43, 55, 56]. The principal question, arising with regard to the radiation
transfer during laser treatment of powders, is the influence of the wavelength, material, particle
size, packing structure and density. The subject was studied experimentally [56], by ray-tracing
modeling [7, 43] and by solving the radiation transfer equation (RTE) [41].
Several techniques for measuring powder absorptivity/reflectivity currently exist [56, 57, 58].
One of the most reliable methods was applied by N.K. Tolochko (so-called radiometric tech-
nique). He used an integrating sphere for direct measurement of a powder absorptivity.
The radiometric technique was applied in LGPP for 18Ni(300) Maraging Steel powder [59].
Experiments showed that the effective absorptivity of the powder is quite high compare to the
absorptivity of the bulk material: αexpp = 0.7−0.8.
These experimental results are confirmed by the solution of the radiation transfer equation
(RTE), which was performed by A.V. Gusarov in [41]. In his method he applies the statistical
scattering model to a powder bed. In the model radiative properties in packed powder beds
of non-spherical particles with arbitrary orientation (statistically isotropic powder bed) are
expressed as the properties of packed beds of spherical particles with the same porosity and
specific surface.
According to the Gusarov’s solution of RTE the effective absorptivity of 18Ni(300) Maraging
Steel powder bed is in the range αtheorp = 0.78−0.88. This theoretical estimation is in good
correspondence with the results obtained by means of the integrating sphere αexpp . In our
simulation we will use the value αp= 0.8 for the effective absorptivity of 18Ni(300) Maraging
Steel powder.
19
Chapter 3. 18Ni(300) Maraging Steel Powder
Figure 3.5 – Discrete thermal resistances model for powder beds by A.V. Gusarov (pic. [3]).
Experiments for the definition of the optical penetration depth were also performed by means
of the integrating sphere [59, 60]. According to the results, the penetration depth of 18Ni(300)
Maraging Steel loose powder is δ' 20 µm.
3.3.2 Powder density
The value of 18Ni(300) Maraging Steel powder density ρp can vary from 50% to 70% of the
density ρs of the solid material. In most of our simulations we will consider the case of a very
loose powder: we assume that the powder density ρp is 50% of the density of the bulk material
ρp = 0.5ρs = 0.004 g/mm3 (see Tab. 3.2).
3.3.3 Effective thermal conductivity
In the framework of the homogeneous medium hypothesis we usually use the concept of
effective thermal conductivity for powders. The effective thermal conductivity relates to a
certain volume (greater than the size of inhomogeneity) of the material and describes the
heat conduction of this volume as a whole. A powder bed is considered as a multiphase
heterogeneous medium, where the solid particles are dispersed in a continuous gas phase in
such a way that they can touch each other and form point contacts between them. Therefore,
the effective thermal conductivity of a powder depends both on the properties of the solid and
of the gas phases.
The model of effective thermal conductivity developed by A.V. Gusarov [3] takes into account
all the types of thermal interaction between particles in powder and packed beds. This model,
applied to packed equal spheres, estimates the effective thermal conductivity as a function of
the volume fraction of solid fs , the mean coordination number N , and the thermal resistance
of a contact between two neighbouring particles Σp .
20
3.3. Powder properties
Powder type d0 fs kp (eq.(3.7)) kp (exp.)
18Ni(300) Maraging Steel 8.87 µm 50−55% 0.0982 W/m/C 0.0847 W/m/C
Commercially Pure Silver 5 µm 60% 0.0857 W/m/C 0.0949 W/m/C
Table 3.3 – Thermal conductivity of powders in air at normal conditions: theoretical predictionand experimental results.
The contact resistance Σp depends on four dimensionless parameters: the relative size of the
contact rc /r0 between powder grains, the ratio of thermal conductivities of the solid and gas
phases ks/kg , the Knudsen number Kn, and the adiabatic exponent of the gas phase γa .
If the particles of an heterogeneous medium form only point contacts (as in the case of loose
powders), the heat flux through the solid phase is negligible and the heat transfer is performed
through the gas gap between powder particles.
For point contacts rc ' 0, at high ks/kg , as it is in a powder bed:
kp
kg= fs N
2
[1
2ln(1+L)+ ln(1+
pL)+ 1
1+pL−1
]
L = 3
4pπKn
γa +1
9γa −5, Kn = l0/d0,
(3.7)
where l0 is a gas mean-free path, d0 - powder grain diameter.
In this study k is the notation of the effective thermal conductivity of the material. In order to
denote the effective thermal conductivities of the powder and of the solid material we use kp
and ks , respectively.
For the validation of the theory, described above, an experimental setup for powder conduc-
tivity measurements, based on the classical flash method [61], was developed [59]. The first
experimental results for several powder types are presented in Tab. 3.3. They are compared to
theoretical prediction of Gusarov’s theory for point contacts (3.7).
In the theoretical estimation of the effective thermal conductivity of powders we assume that
air plays the role of the gas phase between powder particles: γa = 7/5, kg= 0.026 W/m/C,
l0 = 58 nm [3]. The coordination number N = 5.57 corresponds to a randomly packed powder
bed (see [62]).
The results in Tab. 3.3 show that the effective thermal conductivity of a powder is of the order
of several times the conductivity of air and depends on the particles diameter. One can see
good correlation between theory and practice for maraging steel and pure silver powders.
In the simulation we will use the value kp = 0.1 W/m/C for the effective thermal conductivity
of 18Ni(300) Maraging Steel powder.
21
Chapter 3. 18Ni(300) Maraging Steel Powder
3.4 Molten material properties
In this study we assume that the thermal conductivity kl of the liquid 18Ni(300) Maraging Steel
is similar to the value ks for the solid material and is not dependent on temperature. According
to [63], the thermal expansion coefficient for different grades of steels at the melting point
does not exceed 2 ·10−5 1/C. Therefore the density ρl of molten 18Ni(300) Maraging Steel is
approximately equal to the density ρs of the solid material. However the same assumption
cannot be made for the material surface absorptivity, which evolves significantly during
melting.
According to [56, 64, 65] absorptivity of a flat solid stainless steel surface for a wavelength close
to 1 µm at the room temperature is in the range 0.4−0.6, depending on the roughness of the
surface. Whereas the surface absorptivity of liquid steel can vary from 0.1 to 0.3 for the same
wavelength [64, 65].
In this work we choose the values αl= 0.2 and αs= 0.5 for the surface absorptivity of molten
and solid 18Ni(300) Maraging Steel, respectively.
Remark 3.1. Actually the surface absorptivity of molten steel strongly depends on its chemical
composition and on the interaction with the surrounding atmosphere. For example, oxidation of
the liquid material surface can cause a significant increase of absorptivity: up to 0.7, according
to [64, 66]. However in this study we assume that the surface oxidation is negligible. This
assumption is justified since the SLS/SLM process is performed under a protective atmosphere
(argon Ar).
22
3.5. Operation parameters used for Maraging Steel Powder
Laser wavelength λ 1070 nm
Laser beam profile top-hat/gaussian
Laser beam radius ω 40 µm
Laser power P 100; 200; 300 W
Scanning speed υ 1; 2; 3 m/s
Table 3.4 – Process parameters used in the study.
3.5 Operation parameters used for Maraging Steel Powder
The typical process parameters, used in this study, are presented in Tab. 3.4. We have chosen
them in accordance with modern industrial requirements (see Chapter 2). The parameters
form 9 sets (see Tab. 3.5). However some of them will not give good results for the chosen laser
and material parameters. We will discuss it in next chapters.
The choice of the scanning strategy and of the dimensions of the powder layer will be discussed
in each chapter separately.
Set No. Power/speed, W/(m/s)
1 100/12 100/23 100/34 200/15 200/26 200/37 300/18 300/29 300/3
Table 3.5 – Sets of the process parameters used in simulation.
23
Chapter 3. 18Ni(300) Maraging Steel Powder
Parameter Powder bed Solid material Liquid material
Grain radius r0 4.435 µm − −Optical penetration depth δ 20 µm − −Absorptivity α 0.8 0.5 0.2
Thermal conductivity k 0.1 W/m/C 26 W/m/C 26 W/m/C
Density ρ 0.004 g/mm3 0.008 g/mm3 0.008 g/mm3
Table 3.6 – Properties of 18Ni(300) Maraging Steel.
3.6 Conclusions
Tab. 3.6 summarizes the properties of 18Ni(300) Maraging Steel powder, solid and liquid
material, which we will use in the study.
It can be seen that, during sintering/melting of 18Ni(300) Maraging Steel, the thermal con-
ductivity and the absorptivity of the material change drastically. Thermal conductivity of
the powder is of the order of several times the conductivity of air but it is less than 1% of the
conductivity of bulk material. At the same time the liquid material (as all liquid metals) is
highly reflective, while the powder bed effectively absorbs laser energy. In the next chapters,
we will discuss the influence of these phenomena on the SLS/SLM process.
Observe that such a significant change in the properties due to melting is typical for the
majority of metallic powders used in SLS/SLM.
24
4 Single Grain Model
In this Chapter we consider the mechanisms of heating and melting of a separate and perfectly
spherical powder grain irradiated by a plane electromagnetic wave. This theory is often
called Single Grain Model (SGM) [38, 67]. We improve the existing model by considering the
real intensity distribution instead of the homogeneous one and by taking the finite thermal
conductivity of a powder particle into account.
The main outputs of the Single Grain Model will be:
• an estimation of the melting time of a grain;
• a relatively precise information about the evolution of the average temperature inside a
grain.
These results will be used later on to build the reliable consolidation model we need to
integrate into the macroscopic description of the SLS/SLM process.
25
Chapter 4. Single Grain Model
Figure 4.1 – Irradiation of the surface of a powder bed by a moving laser.
4.1 Single Grain Model: main assumptions
Evolution of the effective parameters of a powder bed depends on the melting dynamics of
powder grains and their interaction and rearrangement. A Single Grain Model allows to study
the process of melting of a separate powder grain on the surface of a powder bed during
laser heating. Knowing the melting behavior of a single grain we can better understand the
evolution mechanisms of powder properties during laser sintering/melting.
The Single Grain Model considers powder-laser interaction at a scale of a powder grain size.
The grain is assumed to be spherical and there is no influence of the neighbouring grains.
Remark 4.1. In [38, 67] the grain was assumed to be homogeneously irradiated all over its
surface. Since homogeneous irradiation is difficult to justify, in this study we consider that the
powder particle interacts with a monochromatic polarized plane wave (see Fig. 4.1), which is a
rough approximation of a top-hat/gaussian laser beam. However we will keep the name ‘Single
Grain Model’ for this theory.
In a plane electromagnetic wave, linearly polarized in the direction of x-axis and propagating
in the void along z-axis (see Fig. 4.1), the electric field at a moment of time t = 0 in a point
(x, y, z) is:
E(x, y, z, t ) = E0e−i (kλz−ω f t ) = E0e−i (kλz−ω f t )ex . (4.1)
The wave is characterized by its electric amplitude E0, the wave number kλ = 2π/λ and the
angular frequency ω f = kλc with c the speed of the light.
26
4.1. Single Grain Model: main assumptions
Laser wavelength λ 1070 nm
Laser beam radius ω 40 µm
Laser beam profile top-hat/gaussian
Radius of the grain r0 4.435 µm
Density ρs 0.008 g/mm3
Thermal conductivity ks 0.026 W/mm/C
Dielectric constant εR+iεI (3.24− i 4.34)2
Latent heat of fusion L f 247 J/g
Table 4.1 – Properties of the laser and of a typical 18Ni(300) Maraging Steel powder grain.
4.1.1 Intensity of the plane wave and laser power
An important characteristic of an electromagnetic plane wave is its time-average intensity I0.
This field proves to be constant in space and oriented in the direction of the wave propagation:
I0(x, y, z) = I0ez . (4.2)
For the plane wave (4.1), the connection between I0 and the electric amplitude is:
I0 = kλ2µpω f
E 20 , (4.3)
where µp= 4π ·10−7 N/A2 is the vacuum magnetic permeability.
In the Single Grain Model the plane wave we select to approach the laser beam, has a time-
average intensity I0 equal to the maximum laser beam intensity:
I0 = P
πω2 . (4.4)
for the top-hat or gaussian beam. We use the expression (4.4) both for the top-hat and gaussian
beam, as far as we assume that powder particles we are interested in, i.e. the particles which
are molten, are situated in the center of the laser beam.
27
Chapter 4. Single Grain Model
r
z (polar axis)e.-m. wave
polarisation
Figure 4.2 – Meshing of a 2D domain for the Single Grain Model calculations.
4.1.2 Heating of a spherical grain with a plane wave
The heating by a plane electromagnetic wave of amplitude E0 of a non-magnetic spherical
particle with a complex dielectric constant εd= εR + iεI , can be described by Mie theory [68].
It allows to express the electric field, transmitted to the sphere, in the form:
E = ‖E0‖e, (4.5)
where e is the field transmitted for an incident wave with unit electric amplitude.
Using Mie solution of Maxwell equations, we can calculate the heat source q induced by Joule
heating in the irradiated powder grain. It is proportional to the intensity I0:
q = 1
δI0 ∥ e ∥2 . (4.6)
The factor
δ=− 1
εI kλ(4.7)
in (4.6) has the dimension of a length and corresponds to the optical penetration depth of the
grain.
As we discussed in § 3.3.1 of Chapter 3, penetration depths of bulk metals are relatively small
(tens of nanometers) compare to the radius of the grain (r0 = 4.435 µm). Therefore the volume
source (4.6) is concentrated near its surface and a finite element simulation of the heating
process would require a very fine meshing of the grain. The classical solution to avoid this
issue is to replace the volume heat source (4.6) by an equivalent surface heat source qsurf:
qsurf =−I0n ·st. (4.8)
28
4.2. Melting of a powder grain
Figure 4.3 – Temperature distribution inside a spherical grain after 6 µs of irradiation by theplane wave with the intensity corresponding to the 100 W laser (Tab. 4.1).
In (4.8) st denotes the Poynting vector, transmitted to the sphere in the case of an incident
plane wave with unit intensity.
In the case of a linearly polarized wave, the heat source has a simple dependency with respect
to the polar angleϕ (Fig. 4.2). If the polar axis is aligned with the direction of wave propagation
(z-axis in our case), we can write that:
qsurf(r, z,ϕ) = I0(b0(r, z)+b1(r, z)cos2ϕ), (4.9)
where (r, z,ϕ) are the cylindrical coordinates and where b0 and b1 are functions of r and z only.
The coefficients b0 and b1 can be calculated numerically by means of a simulation tool [69].
4.2 Melting of a powder grain
Using the theory presented above, we can model numerically the melting process of a separate
spherical powder grain of radius r0 irradiated by a laser beam of intensity I0. The properties of
the grain correspond to a typical particle in a 18Ni(300) Maraging Steel powder bed. They are
listed in Tab. 4.1. The dielectric constant εd corresponds to the value for steel [53].
The mesh used to discretize the grain surface is presented in Fig. 4.2 in polar symmetry. In
this context we can study the evolution of the temperature field in the particle and its melting
during irradiation with a standard finite element method.
29
Chapter 4. Single Grain Model
Time, µs
Pro
po
rtio
no
fliq
uid
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.00.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
100 W
200 W
300 W
Figure 4.4 – Evolution of molten material proportion during grain melting by a plane wave ofdifferent intensities approaching laser beam with power 100 W, 200 W, 300 W.
Fig. 4.3 presents the temperature distribution inside a typical Maraging Steel powder grain
after 6 µs of irradiation by a plane wave (λ= 1070 nm) approaching a laser beam with 100 W
power (for the detailed laser and grain properties see Tab. 4.1).
Different simulations are performed for plane wave intensities corresponding to laser powers
100 W, 200 W and 300 W. The latent heat of fusion L f is taken into account (see Tab. 4.1).
Evolution of the proportion of molten material during melting for these parameters is shown
in Fig. 4.4. As it can be seen from this graph, the grain, irradiated by the plane wave, is not
molten immediately.
We can calculate the time necessary to initiate melting from the beginning of the grain irradia-
tion by the wave. We denote this time as t s (‘solidus’). We also estimate the time t l (‘liquidus’),
after which the grain is completely molten. Results for the intensities, corresponding to the
chosen laser powers, are presented in Tab. 4.2.
Laser power P Wave intensity I0 (4.4) t s t l t imax
100 W 19894 W/mm2 1.84 µs 6.88 µs 24−71 µs
200 W 39789 W/mm2 0.54 µs 4.08 µs 24−71 µs
300 W 59683 W/mm2 0.26 µs 3.1 µs 24−71 µs
Table 4.2 – Single Grain Model. (i) t s : start of the grain melting (Tab. 4.1); (ii) t l : complete grainmelting; (iii) t i
max: maximum time of grain irradiation (Fig. 4.5).
30
4.3. Evolution of powder bed properties
Figure 4.5 – Maximum irradiation of a powder grain by a top-hat laser beam.
From Tab. 4.2 it can be concluded, that for the selected parameters the melting of a typical
18Ni(300) Maraging Steel powder grain starts in a period of time less than 2 µs, and the particle
is completely molten after 3−5 µs.
In Tab. 4.2 the values of t s and t l are compared to the maximum time of grain irradiation t imax.
To define this time we go back to the system ‘powder-laser’. Consider a top-hat laser beam of
radius ω, travelling with a speed υ and irradiating a powder bed. We assume that the powder
bed consists of equally spherical particles of radius r0. The maximum time while a powder
grain can be located entirely within the border of the laser spot (see Fig. 4.5) is:
t imax = 2(ω− r0)/υ. (4.10)
According to (4.10), for laser scanning speeds 1−3 m/s, t imax is 24−71 µs (Tab. 4.2).
From Tab. 4.2 we can see, that the time of the particle complete melting is several times less
than its irradiation time by the laser.
4.3 Evolution of powder bed properties
We have shown before, that the irradiation of a powder grain by a plane wave is not homoge-
neous. The example of the temperature gradient inside the grain can be seen in Fig. 4.3. From
this picture it is clear that when a part of the grain is already molten, a part of it still stays solid.
At the level of a powder bed, when a powder grain begins to melt, the molten material starts
to form necks with the neighbouring particles. As it is shown in [1], at the first moment
necks between particles are small and unstable, while during further laser irradiation they
are merging and growing. Formation of stable interparticular necks between powder grains
can be considered as a starting point of the evolution of the effective properties (absorptivity,
thermal conductivity) of the powder bed.
31
Chapter 4. Single Grain Model
Figure 4.6 – Evolution of the proportion of the molten material during irradiation of twoneighbouring grains: (i) t < t 0; (ii) t ' t 0, (iii) t > t 0.
To connect the melting of a separate powder grain to the properties of a powder bed, i.e. to
link the Single Grain Model to the homogeneous medium hypothesis, commonly used for
SLS/SLM simulation (see Chapter 6 for details), we will estimate the evolution of the average
temperature of the grain during its melting by a plane wave.
We need to define the moment when a neck starts to form between two neighbouring powder
grains. For this purpose we will make some assumptions on the irradiated powder bed:
• Surface of the powder bed is perfectly ‘flat’, i.e. the centers of powder grains in the
surface layer of the powder are situated in the same horizontal plane.
• Two neighbouring grains on the surface of the powder bed are irradiated in the same
way by the laser.
• Back-scattering of laser irradiation by powder grains is not taken into account.
• Heat transfer between the grains during their melting is not taken into account.
The position of a contact point between two neighbouring grains is shown in Fig. 4.1. During
laser irradiation of the grains, the time t 0, when the melt front reaches the contact point, can
be considered as the time of the neck-forming, i.e. when powder bed properties evolution
starts.
We use the Single Grain Model to estimate the value of t 0. We study the temperature evolution
in a point on the surface of a spherical particle, irradiated by a plane wave. The particle is made
of 18Ni(300) Maraging Steel (see properties in Tabs. 4.1 and 3.6), and the point we consider is
a contact point to a neighbouring particle (see Fig. 4.1).
32
4.3. Evolution of powder bed properties
Time, µs
Tem
per
atu
re,
C
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.00.0
500
1000
1500
2000
2500
3000
t 01t 0
2t 03
100 W
200 W
300 W
Figure 4.7 – Temperature evolution in the contact point between two grains irradiated by aplane wave, approaching the laser of different powers.
The intensities of the plane are adapted to approach a top-hat laser beam with a radius 40
µm and powers 100 W, 200 W or 300 W (Tab. 4.3). The simulated temperature evolution is
presented in Fig. 4.7. The time t 0, when the melt front reaches the contact point, is also shown
on this figure and its exact value is presented in Tab. 4.3.
Fig. 4.6 presents the stages of the evolution of the molten material proportion during laser
irradiation of two neighbouring grains on the surface of a powder bed. As we can see, at the
moment, when the melt front reaches the contact point between the particles, part of each
particle still stays solid. To connect these results to the homogeneous medium model, we will
estimate the average temperature of the grains T 0av at the moment t 0.
In Fig. 4.8 one can observe the evolution of the average temperature of the grains T 0av during
their irradiation with a plane wave. The exact values of T 0av for the chosen intensities/powers
are presented in Tab. 4.3.
Laser power P Wave intensity I0 (4.4) C. p. melt time t 0 Av. temp. T 0av
100 W 19894 W/mm2 3.92 µs 1353 C
200 W 39789 W/mm2 1.76 µs 1213 C
300 W 59683 W/mm2 1.04 µs 1089 C
Table 4.3 – Melting time t 0 of the contact point between two grains, irradiated by a plane waveof intensity I0, and the average temperature T 0
av of the grains at this moment.
33
Chapter 4. Single Grain Model
Time, µs
Ave
rage
tem
per
atu
re,
C
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.00.0
500
1000
1500
2000
2500
3000
t 01 t l
1t s
1
T f
100 W
200 W
300 W
Figure 4.8 – Evolution of the average temperature of a powder grain (Tab. 4.1), irradiated by aplane wave, approaching the laser of different powers.
As it can be seen from Tab. 4.3, for the selected parameters, the melt front reaches the contact
point, when the average temperature of the grain T 0av is less than the melting point T f = 1413C
of the material.
It has to be observed on Fig. 4.8 that the time dependency of the average temperature of the
grain T 0av, heated by a plane wave, is not linear. The non-linearity is mostly explained by the
phase transition. For example, if we look at the curve, corresponding to a laser power of 100 W,
we see that the non-linear part is enclosed between t s1 = 1.84 µs and t l
1 = 6.88 µs. These values
correspond to the moments when the grain starts to melt and when it is completely molten,
respectively (see Tab. 4.2).
Outside the interval (t s , t l ), Tav depends almost linearly on time. Since we neglect losses for
the surface of the grain, the slope of the graph is given by the ratio between the absorbed
power and the (almost constant) thermal mass of the grain Mth
Mth = 4
3πr 3
0ρsCp . (4.11)
In this formula Cp is the heat capacity of 18Ni(300) Maraging Steel: Cp ' 0.5 J/g/C for any
temperature (see Fig. 3.3).
34
4.4. Different powder grain sizes
4.4 Different powder grain sizes
Fig. 3.1 shows, that the particles in the 18Ni(300) Maraging Steel powder we study are actually
not of the same size. It is therefore important to discuss the value of the time to contact point
melting t 0 and the average temperature T 0av for different radii of powder grains. According to
Fig. 3.1, besides r0 = 4.435 µm, we chose values 2.5 µm and 10 µm.
Fig. 4.9a presents the values of time t 0 for the grains of the three chosen radii, irradiated by a
plane wave, approaching a laser beam of waist 40 µm. We select laser power range from 50 W
to 400 W.
It can be seen from Fig. 4.9a, that for the laser powers 100−300 W, even for the maximum value
of the grain radius (10 µm), the time t 0 is always less than 10 µs. As we already did before, we
are going to estimate the average temperature T 0av of the grains at the moment of time t 0. The
results are shown in Fig. 4.9b.
From Fig. 4.9b one can observe, that for laser powers between 100 and 400 W, the average
temperature of two neighbouring grains stays significantly below the melting temperature T f
at the moment when they start to be connected by a neck. For higher laser powers the surface
melting of a particle is faster, therefore the average temperature of the grain at the moment,
when powder sintering begins, will even be lower.
In the chosen range of the process (Tab. 3.4) and material (Tab. 3.6) parameters, the conclusion
is that the homogenized temperature, at which sintering process starts, is significantly lower,
than the melting point T f of the material.
35
Chapter 4. Single Grain Model
Laser power, W
Tim
et0
,µs
0.0 50 100 150 200 250 300 350 400 4500.0
4.0
8.0
12.0
16.0
20.0
r0 = 2.5 µm
r0 = 4.435 µm
r0 = 10 µm
(a) Time to the contact spot melting t 0 between two equal spherical grains for different grain radii r0.
Laser power, W
Ave
rage
tem
per
atu
reT
0 av,
C
0.0 50 100 150 200 250 300 350 400 4500.0
300
600
900
1200
1500
T f
r0 = 2.5 µm
r0 = 4.435 µm
r0 = 10 µm
(b) Average temperature T 0av of a powder grain at the moment t 0 for different grain radii r0.
Figure 4.9 – Single Grain Model: Time to the contact spot melting t 0 and average temperatureT 0
av of the powder grain at the moment t 0.
36
4.5. Conclusions
4.5 Conclusions
In this chapter we considered the melting process of a separate particle in the surface layer of
a powder bed. The Single Grain Model has been applied for the particular case of a standard
18Ni(300) Maraging Steel powder (Tab. 3.6). For typical process parameters (Tab. 3.4) the
following results have been obtained:
(i) Heating and complete melting of a powder grain on the surface of a powder bed under
a laser takes less than 10 µs from the beginning of the irradiation. For the chosen
parameters this time equals to 14−40% of the maximum irradiation time.
(ii) The time, when the melt front reaches the contact point between two similarly irradiated
neighbouring powder grains on the surface layer of the powder bed, can be considered
as the start of neck-forming and the beginning of the evolution of powder bed effective
parameters. In our case, this process starts less than 4 µs after the beginning of the
irradiation. This delay is between 5−16% of the maximum irradiation time.
(iii) The average temperature of a grain at the moment, when it is molten enough to form
a neck with its neighbour, is less than the melting temperature T f of the material. It
can be said that evolution of the properties of the surface of a powder bed during
laser scanning, at the scale of powder grain size, starts before the moment when the
homogenized temperature reaches the value T f . For the chosen range of laser power
(100−300 W) the evolution process starts at an average temperature of 1000−1300C.
However this threshold value depends on laser power and powder material.
37
5 Melting Dynamics
As we already discussed in Chapter 4, in SLS/SLM when powder grains, irradiated by a laser,
start to melt, the molten material forms necks between them. The formation of necks is
caused by the action of capillary forces and strongly depends on dynamic viscosity and surface
tension of the liquid phase. The process of neck-forming between powder grains defines the
evolution of effective powder properties during laser sintering.
In this Chapter we study capillary parameters of 18Ni(300) Maraging Steel under SLM con-
ditions and their influence on the sintering process. We also discuss the mechanisms of
interconnection of powder grains during their melting. We propose a model which considers
the interaction of molten powder particles as a coalescence of liquid drops.
39
Chapter 5. Melting Dynamics
Temperature, C
Dyn
amic
visc
osi
ty,m
Pa·s
1538 1600 1700 1800 1900 2000 2100 2200 2300 2400 25000.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.0
10.0
Figure 5.1 – The dynamic viscosity of pure liquid iron in the range of temperatures1538C < T < 2500C [4].
5.1 Capillary properties of molten powder material
Rearrangement, interconnection and merging of powder particles during laser sintering/melt-
ing is driven by the capillary motion of the molten material. Description of the capillary flow
of any material is based on its surface tension and dynamic viscosity. The capillary parame-
ters are strongly dependent on the material properties and on the conditions of the melting
process. In this Section, using 18Ni(300) Maraging Steel powder as an example, we discuss the
main factors which can influence the capillary properties of a molten material during SLM of
a metallic powder.
5.1.1 Dynamic viscosity
Viscosity is a property arising from collisions between neighbouring particles that are moving
at different velocities. Dynamic viscosity of a liquid expresses its resistance to shearing flows.
In this study we denote it as µ. Main factors influencing the dynamic viscosity are the chemical
composition of the liquid and its temperature.
An empirical formula for the dynamic viscosity of a liquid metal which takes its chemical
composition into account, is proposed in [70]. According to this paper, dynamic viscosities of
different grades of liquid steel can vary from 6 to 20 mPa·s near the melting point of a material.
The Arrhenius equation is often used to describe the dependence on temperature of the
dynamic viscosity of a liquid [71]:
µ(T ) =µ0eEa /Rg (T+273). (5.1)
40
5.1. Capillary properties of molten powder material
In (5.1) µ0 is a coefficient, Rg= 8.314 J/mol/K is the universal gas constant and Ea is the
activation energy. It corresponds to the potential energy barrier for the movement of an atom
created by the nearest atoms in the liquid.
In [4] the coefficient µ0 and the activation energy Ea are found empirically for liquid iron in
the range of temperatures 1538C < T < 2500C:
µ0 = 0.253 mPa·s, Ea = 47.7 kJ/mol. (5.2)
In Fig. 5.1 we can observe the dependence of the dynamic viscosity µ of liquid pure iron on
temperature T in the range 1538C < T < 2500C. The first value of this range is the melting
point of iron. It can be seen, that the dynamic viscosity strongly depends on temperature
and decreases significantly (from about 6 to 2 mPa·s) with temperature increase, which can
influence interconnection and rearrangement processes of powder particles during SLM. We
will discuss this question in this Chapter.
5.1.2 Surface tension
The surface tension of a fluid is a direct measure of the intermolecular forces acting at the
surface and can be defined as the force per unit length acting parallel to the free surface.
Data are sometimes expressed in terms of free surface energy, because surface tension also
represents the free energy per unit surface. In this work the surface tension of the liquid-gas
interface will be denoted as γ.
The surface tension of pure liquid metals depends on the strength of the cohesive forces
acting between neighbouring atoms, and it can be related to physicochemical properties of
the material. The principal parameters affecting the surface tension of a liquid-gas interface
are the chemical composition of the liquid and the gas, the temperature and the chemical
reactions with a surrounding atmosphere. However, in this study, it will be assumed that the
molten material does not react with the protective atmosphere under which SLM process is
performed.
In a first approximation, the influence of the temperature of a liquid on its surface tension can
be described by a linear law [72]:
γ(T ) = γ f +γ′(T −T f ). (5.3)
In (5.3) γ f is the surface tension of the metal at the fusion point T f and γ′ = dγ/dT is so-called
temperature coefficient of the surface tension.
In [72] γ′ for pure liquid metals is calculated on the basis of the nearest-neighbour interaction-
broken-bond model. The valueγ′ =−0.23·10−3 J/m2/K for pure iron is obtained and confirmed
experimentally.
41
Chapter 5. Melting Dynamics
Parameter 1538C 2500C
Surface tension γ 1.79 N/m 1.51 N/m
Dynamic viscosity µ 6 mPa·s 2 mPa·sDensity ρl 0.008 g/mm3 0.008 g/mm3
Table 5.1 – Properties of liquid 18Ni(300) Maraging Steel, used in Chapter 5.
We use γ f ' 1.79 J/m2 as the initial value of the surface tension in (5.3). It corresponds to the
surface tension of liquid sulfur free stainless steel at temperature about 1500C in a pure argon
atmosphere [73].
One can see from (5.3), that for pure iron the temperature coefficient of the surface tension
γ′ is quite small. We will use the same range of temperature T as for the dynamic viscosity µ
(Fig. 5.1): in the range 1538C < T < 2500C the surface tension of liquid iron decreases only
from 1.79 to 1.51 J/m2.
The surface tension of a material also depends on its chemical composition. According to [73],
the surface tension γ f of steel at its melting point can vary from 1.6 to 1.9 J/m2, depending on
its chemical composition. For various types of steel the character of the dependence of the
surface tension on temperature can also be different. However in [73] we observe a similar
behavior (slight decrease while the temperature is increasing) for several grades of steel.
5.1.3 Summary
In Tab. 5.1 we summarized the parameters of 18Ni(300) Maraging Steel which will be used in
this Chapter.
For the estimation of the temperature influence on the process of powder melting, we have
chosen the range of temperatures 1538C < T < 2500C. It approximately corresponds to the
temperatures between the fusion point and the vaporization point of 18Ni(300) Maraging
Steel: T f = 1413C, Tv = 2862C. Using laws (5.1) and (5.3), we calculated the values of the
surface tension and the dynamic viscosity for the minimum and maximum temperatures in
this interval. We also assume that the density of the liquid material ρl does not depend on
temperature (see § 3.4 of Chapter 3).
42
5.2. Liquid drops coalescence model
Figure 5.2 – Coalescence of two liquid drops.
5.2 Liquid drops coalescence model
When powder grains start to melt, they get connected by small necks [1]. Therefore contact
points between powder particles are transformed to contact spots, which significantly changes
heat transfer mechanisms in a powder bed.
In Chapter 4 we already discussed the starting time of effective powder bed properties evolu-
tion in the framework of the Single Grain Model. It was assumed that the neck-forming begins
when the melt front in a powder grain, irradiated by a laser, reaches the point of contact with
a neighbouring grain. We considered 18Ni(300) Maraging Steel powder and typical process
parameters (Tabs. 3.5 and 4.1): laser powers 100−300 W, scanning speeds 1−3 m/s. We have
shown that for typical powder particles (r0 = 4.435 µm) the time when the neck starts to form,
is 1−4 µs after the beginning of laser irradiation and that the particles are completely molten
in 3−7 µs.
This theory gives us an idea of the order of magnitude for the molten material lifetime, during
which necks form between grains. We believe that the neck-forming mechanism is essentially
driven by capillary forces. To prove that gravity plays no role, we have to consider the Bond
number [74]. It compares the surface tension to the gravitational forces:
Bo = ∆ρg L2c
γ, (5.4)
In (5.4) ∆ρ = ρl −ρatm is the difference in densities of the liquid and gas phases, g – the
gravitational acceleration, Lc – the characteristic length of the liquid volume, and γ is the
liquid-gas surface tension. For a spherical liquid drop the characteristic length Lc is equal
to the grain size d0 = 8.87 µm. Argon is chosen as a gas phase: ρatm = 1.65 ·10−6 g/mm3. For
18Ni(300) Maraging Steel (ρl = 0.008 g/mm3) the Bond number Bo is small:
Bo ' 3.45 ·10−6 ¿ 1. (5.5)
43
Chapter 5. Melting Dynamics
Drop radius r0 Temperature T Parameter K2
2.5 µm 1538C 0.554.435 µm 1538C 0.584.435 µm 2500C 0.6610 µm 1538C 0.6
Table 5.2 – Parameter K2 (see eq.(5.7)) for different temperatures and radii of the drops [5].
Thus the behavior of a molten powder grain is driven by surface tension forces only: the
gravitational forces are negligible and the molten grain keeps its spherical form in the absence
of capillary interactions with neighbouring grains (or a substrate). Therefore we can assume
that the interaction between two molten particles with an initial contact point between them
can be described as a coalescence of two spherical drops.
Coalescence of liquid drops was studied by Frenkel [75]. He proposed a law for the time
evolution of the contact radius rc between two equal spherical liquid drops (Fig. 5.2):
rc (t ) = K1
(r0γ
µ
)1/2
t 1/2, (5.6)
where K1 is a unitless geometry-dependent calibration coefficient, r0 – the radius of the drop,
γ and µ are the surface tension and the dynamic viscosity of the liquid material, respectively.
A similar law (rc ∼ t 1/2) was proposed in [76]. However it was only valid for the initial stage of
the coalescence: rc ¿ r0 (in [76] it was used for rc /r0 < 0.03).
Menchaca-Roca in [77] proposed a power law rc ∼ tα0 for the description of the entire process
(until rc ' r0): α0 = 0.55 for the beginning of the coalescence and α0 = 0.41 for larger neck
radii between liquid drops.
Thoroddsen in [5] presented a differential equation describing the velocity of a neck growth
between two equal spherical liquid drops:
drc
dt= K2
rc
√γ
2ρl
(r0 −2rc +
√r 2
0 − r 2c
)1/2, (5.7)
where K2 is a unitless proportionality constant. The choice of K2 depends on the value of
ρl r0γ/µ2 (see [5] for the details), i.e. on the properties of the liquid and on the radii of the
drops.
The surface tension γ, the dynamic viscosity µ and the density ρl of liquid 18Ni(300) Maraging
Steel are listed in Tab. 5.1. According to these data we can set the values of the parameter K2
for the Thoroddsen’s equation. They are presented in Tab. 5.2.
44
5.2. Liquid drops coalescence model
Time, µs
r c/r
0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.00.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Lagrangian FE sim.
Thorrodsen eq. (5.7)
Shikhmurzaev eq. (5.9)
Figure 5.3 – Relative contact spot radius evolution: simulation and theoretical results,r0 = 4.435 µm
Remark 5.1. Equation (5.7) is solved using MATLAB software with the initial condition for the
contact spot radius
rc (0) = 10−3 µm, (5.8)
rc (0) = 0 is not recommended due to the form of the right-hand side of (5.7).
Later a similar power law for the evolution of a contact spot between two drops with initial
point contact was obtained by Shikhmurzaev [78]:
rc (t ) = K3
(2r0γ
ρlτM
)1/2
t 1/2, (5.9)
where τM – Maxwell relaxation time of the liquid. In the sequel we will use the Maxwell
relaxation time of steel τM = 10−15 s [79]. In (5.9) K3 is a phenomenological constant related
to the inverse compressibility of the liquid. It is expressed with a unit of s/m.
The similarity of laws (5.6) and (5.9) seems to be promising, as far as they are based on different
ideas: Frenkel (5.6) considers the coalescence of two drops from the energetic point of view,
while Shikhmurzaev’s hypothesis (5.9) is based on the motion of wetting lines and material
fluxes from one interface to another through these lines. However (5.9) requires additional
analysis of liquid material.
45
Chapter 5. Melting Dynamics
Time, µs
r c/r
0
0.0 0.5 1.0 1.5 2.0 2.4 3.00.0
0.2
0.4
0.6
0.8
1.0
r0 = 2.5 µm
r0 = 4.435 µm
r0 = 10 µm
Figure 5.4 – Relative contact spot radius evolution for different radii of liquid drops: solid lines– LFEM, dashed lines – Shikhmurzaev’s eq. (5.9) and Thorrodsen’s eq. (5.7).
In order to estimate K3 for the liquid parameters of liquid, presented in Tab. 5.1, coalescence
of two drops was modelled by means of the Lagrangian finite element method (LFEM) [80].
Evolution of the relative contact radius rc /r0 in time, obtained numerically1, can be seen in
Fig. 5.3.
There is a clear possibility to fit the Shikhmurzaev law (5.9) with LFEM results at least for
sufficiently large time t (see Fig. 5.3). In case of powder grains with radii r0 = 4.435 µm at
melting temperature T f the best fit is obtained for the value
K3 ' 3.4 ·10−6 s/m (5.10)
of the Shikhmurzaev phenomenological constant.
In Fig. 5.3 we have also added the numerical solution of Thoroddsen’s equation (5.7) for
comparison. We can see, that (5.7) implies rc /r0 ≤ 0.8. However the results are in quite good
correspondence.
r0, µm Temperature T , C tc (rc ' r0), µs
2.5 1538 0.54.435 1538 0.84.435 2500 0.3510 1538 1.9
Table 5.3 – Coalescence time of equal liquid drops.
1The data are provided by Dr. P. Clausen.
46
5.2. Liquid drops coalescence model
Time, µs
r c,µ
m
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.00.0
1.0
2.0
3.0
4.0
5.0
r = 4.435 µm
r = 2.5 µm
2.5 µm – 4.435 µm
4.435 µm – 10 µm
Figure 5.5 – Contact spot radius evolution between two liquid drops of different radii: (i)r1 = 2.5 µm, r2 = 4.435 µm; (ii) r1 = 4.435 µm, r2 = 10 µm (eq. (5.11)).
As we have already observed in Chapter 4, analysing a grain with radius r0 = 4.435 µm is not
enough to get a comprehensive overview of the used 18Ni(300) Maraging Steel powder shown
in Fig. 3.1. We also have to consider the case of different grains, corresponding, for instance,
to extreme radii 2.5 µm and 10 µm. In Fig. 5.4 we present the coalescence results for such
grains as they are obtained from LFEM (solid lines) and eqs. (5.7) – (5.10) (dashed lines). It
can be seen that for two equal liquid maraging steel drops of radii 2−10 µm with an initial
point contact between them, the time tc , when rc ' r0, is in the range between 0.5 and 2 µs.
For grain radii < 5 µm the two hypotheses (5.7) and (5.9) give quite similar results for the
chosen parameters and rc /r0 ≤ 0.8. The values of tc for drops at melting temperature T f can
be estimated from Figs. 5.3 – 5.4. The results are presented in Tab. 5.3.
For typical 18Ni(300) Maraging Steel powder grains (r0 = 4.435 µm) tc does not exceed 1 µs.
According to the results of Chapter 4, this time is much shorter than the time of complete
melting for a particle of radius r0 irradiated by a laser with a power 100−300 W (3−5 µs).
Remark 5.2. Since the surface tension γ and the dynamic viscosity µ are strongly dependent
on temperature (see (5.1) and (5.3)), the contact spot evolution might also be affected. We
can compare the values tc for the minimum and maximum temperatures from the range
1538C < T < 2500C. The case of equal grains of radii r0 = 4.435 µm has been analysed for a
temperature close to vaporization point Tv . The results can also be found in Tab. 5.3. According
to Shikhmurzaev theory (5.9), the increase of temperature provokes the increase of the drops
coalescence speed.
47
Chapter 5. Melting Dynamics
r1, µm r2, µm Temperature T , C tc (rc ' min(r1,r2)), µs
2.5 4.435 1538 ∼ 0.44.435 10 1538 ∼ 0.9
Table 5.4 – Coalescence time of liquid drops with different radii.
The Thoroddsen’s equation (5.7) generalizes to the case, when the coalescence takes place
between two drops of different radii r1 and r2:
drc
dt= K2
√γ
ρl
(1
ς(t )− 1
rc (t )
),
ς(t ) = r1 + r2 −√
r 21 − r 2
c (t )−√
r 22 − r 2
c (t ).
(5.11)
Fig. 5.5 represents the evolution of the contact spot radius between two drops in two different
cases r1 = 2.5µm, r2 = 4.435µm and r1 = 4.435µm, r2 = 10µm. When the grains have different
radii, we define tc as the time when rc ' min(r1,r2). The values of tc , measured on Fig. 5.5, are
listed in Tab. 5.4.
From Figs. 5.4 – 5.5 we conclude that for the chosen radii of liquid drops, the coalescence time
does not exceed 2 µs. For a typical 18Ni(300) Maraging Steel powder grain (4.435 µm) this time
is even less than 1 µs, which is small compare to the corresponding melting time under a laser
(3−5 µs, see Chapter 4).
48
5.3. Conclusions
5.3 Conclusions
In this Chapter we discussed the liquid-gas surface tension and the dynamic viscosity of liquid
18Ni(300) Maraging Steel. We have shown that an increase of temperature provokes a decrease
of these parameters. However, the surface tension decreases slightly, while the reduction of
the dynamic viscosity is significant. Capillary parameters of the material are summarized in
Tab. 5.1.
We proposed a model for the initial stage of powder grains sintering. It considers the coales-
cence of liquid drops. The model describes the time evolution of the contact between two
liquid drops using different ‘square-root laws’ rc ∼ t 1/2. The difficulty is that these laws involve
empirical parameters. They have to be chosen according to the properties of the material
and the geometry of the powder bed. We computed these parameters for liquid 18Ni(300)
Maraging Steel by means of the Lagrangian finite element method.
The coalescence model has been used to show, that for 18Ni(300) Maraging Steel powder
the time to reach merging between two molten particles does not exceed 1 µs. This time is
relatively small compared to the typical time of complete melting of a powder grain (3−5 µs,
see Chapter 4).
We have also shown, that, due to the decrease of the surface tension and the dynamic viscosity,
the temperature increase can even accelerate the process of drops coalescence. Surface
tension and dynamic viscosity do not intervene the coalescence process independently. It
depends certainly only on the ratio γ/µ, which has the dimension of speed.
49
6 Numerical Modelling of SLS/SLM
A finite element simulation software has been developed at LGPP for the modelling of SLS/SLM
processes [38]. This software, in contrast to commercial simulation tools, is built specially for
Selective Laser Sintering and Melting.
In this Chapter, we discuss a numerical model used for the simulation of SLS/SLM. The model
is based on the homogeneous medium hypothesis. We describe its main assumptions and
advantages. In particular, we propose the use of two grids. One very fine mesh to achieve an
accurate description of the geometry of the part under construction and one relatively coarse
– to follow the temperature evolution with as few algorithmic complexity as possible. However
local refinement will be needed to be sensitive, at the scale of the powder bed, to the fine
features of the laser beam.
51
Chapter 6. Numerical Modelling of SLS/SLM
substrate
powder bed
Σ
x
zy
L
l
hsubs.
hpowd.
Figure 6.1 – Typical domainΩ: powder bed (with top surface Σ) on a substrate.
6.1 Numerical model description
The simulation software works in the framework of the homogeneous medium hypothesis. A
powder bed is considered as an homogeneous medium. Material properties of this medium are
characterized by averaged (so-called effective) parameters: absorptivity, thermal conductivity,
etc. The parameters are averaged over a homogenization volume. For powders the dimensions
of the homogenization volume should be several times larger than the characteristic size of
the material, i.e. the size of a powder grain d0.
In the framework of the model, heat transfer in a parallelepipedonΩ representing the pow-
der bed and its immediate surrounding (like the substrate, see Fig. 6.1) is described by the
evolution of two fields: the enthalpy per unit mass u and the temperature T .
6.1.1 Enthalpy and temperature evolution
The evolution of the first unknown, the enthalpy per unit mass u, follows the heat diffusion
equation, completed by boundary and initial conditions [38]:
ρ∂u
∂t−div(k∇T ) = F, inΩ,
−k∂T
∂n= κ(T −T0)+ f , on ∂Ω,
u = u0, inΩ, for t = 0.
(6.1)
52
6.1. Numerical model description
In equations (6.1) k denotes the effective thermal conductivity of the medium, ρ is its density,
κ is the thermal exchange coefficient between the domainΩ and the environment of tempera-
ture T0, F is a (given) volumic heat source, f is a (given) boundary heat source. The outwards
normal derivative of the temperature on the surface ∂Ω is defined by ∂T∂n .
In SLS/SLM, the heat source is volumic, meaning that f is in principle zero. In some cases, the
heat source will have to be approached by a surface source for numerical reason: F = 0 and
f 6= 0 on the top surface Σ: z = hsubs. +hpowd. (see Fig. 6.1).
It has also to be pointed that the thermal exchange coefficient κ is not well known except on
the top surface Σ (see Chapter 8). The uncertainty on κ is however not an issue. The lateral
surfaces and the bottom surface z = 0 (see Fig. 6.1) are (thermally) far away from the laser
activities and the solution will be indifferent to the boundary conditions imposed there.
The equation, which connects the second unknown, the temperature T , with the enthalpy per
unit mass u, is an algebraic relation [38]:
T =β(u), inΩ. (6.2)
The function β, used in (6.2), is a non-decreasing function. Out of phase transitions, it can be
directly connected to the heat capacity of the material Cp , by the formal relationship:
dβ
du= 1
Cp, for Cp =const. (6.3)
The function β also allows to take the energy, consumed or released during phase transforma-
tions of the material, into account.
6.1.2 Sintering potential and effective properties
For a precise description of SLS/SLM process a third field is introduced. This field is called
sintering potential Φ and contains the information about the state of the material during
SLS/SLM [38, 81]. Its value is 0 in loose powder and the value 1 corresponds to fully dense bulk
material.
The effective thermal conductivity k in (6.1) depends on sintering potential: k = k(Φ). During
SLS/SLM the effective thermal conductivity of a medium evolves from the value kp of a loose
powder to the value ks corresponding to a bulk material (see § 3.3.3 of Chapter 3) and can be
connected to the sintering potentialΦ by interpolating between kp and ks :
k(Φ) = (1−Φ)kp +Φks . (6.4)
53
Chapter 6. Numerical Modelling of SLS/SLM
In this work we actually consider equation (6.4) as the definition for the sintering potential
Φ. It respects the conditionΦ= 0 for loose powder andΦ= 1 for bulk material. The sintering
potential might also be used to evaluate another effective property of the powder bed: its
absorptivity (see Chapter 3). The absorptivity α decreases from the high value αp in powder
to the low one αl for liquid materials (see § 3.3.1 of Chapter 3). In the example of 18Ni(300)
Maraging Steel we can see that even the absorptivity of the solid material αs also often differs
from the absorptivity of its liquid state αl (see Tab. 3.6).
In the framework of our numerical model, we deduce α from the value of the sintering
potential by means of an interpolation rule similar to (6.4). However, due to phase transition,
the dependence of α on the local value of temperature can not be neglected
α(Φ,T ) = (1−Φ)αp +Φαbulk(T ) (6.5)
with
αbulk(T ) =αs , if T < T f ,
αl , if T ≥ T f .(6.6)
If the interpolation rule (6.4) for the thermal conductivity is considered as exact and serves as
a definition for the sintering potentialΦ, consider that eq. (6.5) is only an assumption.
6.1.3 Sintering potential evolution
In fact, the sintering potential relates to the connections between particles of a sintered powder.
It evolves in time according to an ordinary differential equation with initial data which can be
written as [75]:
∂Φ
∂t= ζ(1−Φ), inΩ,
Φ= 0, inΩ, for t = 0.(6.7)
where the parameter ζ, called sintering rate, is the inverse of the characteristic sintering time
τsint of the material.
We have developed the phenomenological law, which allows to compute the sintering rate ζ at
any point x of the powder bed at any moment of time t . According to a microscopic model for
the liquid flow dynamics between neighbouring powder particles, we assume:
ζ(x, t ) =1/τsint, if 3
4πδ30
∫B(x,δ0) T (y, t )d3y ≥ Tt
0, otherwise(6.8)
where B(x,δ0) denotes the sphere of radius δ0 centered at the point x. The three parameters:
sintering time τsint, radius δ0 and threshold temperature Tt , involved in the law (6.8), are
connected to the properties of powder and molten material and to the operating parameters.
54
6.2. Numerical method
In particular, the parameter δ0 is related to the geometrical powder properties (packing, grain
size distribution). In the model, we will often assume it to be equal to the size of a typical
powder grain d0.
On the other hand, the default value for the sintering threshold Tt is the material melting
point T f [38]. However, in general, it also depends on the material properties and operating
parameters. The estimation of the sintering threshold temperature in SLS/SLM simulation
will be discussed in Chapter 7.
Finally, when processing of metallic powders with typical particle sizes less than 10 µm and
with high laser powers (see § 2.1.3 of Chapter 2), the characteristic sintering time τsint turns out
to be very small compare to the laser irradiation time (see Chapters 4 and 5), and the evolution
equations (6.7)-(6.8) forΦmight be replaced by the limit case when τsint → 0. This limit case
means that we can neglect the partial sintering stage and that the sintering potential goes to
1 as soon as the average temperature in the ball of radius δ0 has reached the threshold Tt :
Φ(x, t ) =0, if 3
4πδ30
∫B(x,δ0) T (y, s)d3y < Tt , ∀ s < t ,
1, otherwise.(6.9)
6.2 Numerical method
The numerical approach, described by equations (6.1) – (6.9), not only helps to model SLS/SLM
processes with good precision but also allows to estimate the accuracy of a final sintered part.
In this Section we will describe the numerical method, used for its realisation.
6.2.1 Time discretization
A simulation time step τ is chosen for time discretization. It should be sensitive to the evolution
of time- and temperature-dependent material and process parameters. In the case of melting
of a powder with a continuous laser, the scales of this evolution are the time to melting for the
powder surface and the characteristic sintering time τsint.
We assume τsint to be negligibly small (see § 6.1.3 for the details). For fine powders, like
18Ni(300) Maraging Steel (see Chapter 3), and laser powers between 100 W and 300 W the time
to surface melting is less than 5 µs (see Chapter 8). Therefore, for the simulation of SLS/SLM
under these process conditions, we propose:
τ= 1−5 µs. (6.10)
The time step being chosen, we are looking for a series of profiles un :Ω→R, T n :Ω→R and
Φn :Ω→R approaching the real enthalpy per unit mass, temperature and sintering potential
profiles, respectively, at discrete times tn = nτ, n = 0,1,2 . . ..
55
Chapter 6. Numerical Modelling of SLS/SLM
The initial conditions in (6.1) and (6.7) and the algebraic relation (6.2) gives the profiles for
t0 = 0:
u0 = u0, T 0 =β(u0) andΦ0 = 0. (6.11)
The idea is then to deduce recursively un+1,T n+1,Φn+1 from un ,T n ,Φn, n = 0,1,2, . . .. Im-
plicit connections between un+1,T n+1,Φn+1 and un ,T n ,Φn+1 (like backward Euler scheme)
are favorable. They do not impose strict restriction on the time step τ to be used. These re-
strictions, known as stability conditions [82], depend on spatial discretization and increase
considerably the algorithmic complexity of the numerical method.
The simple profile to deal with is the sintering potential. Applied to (6.7) the backward Euler
scheme gives
Φn+1 = Φn +τζn+1
1+τζn+1 (6.12)
with ζn+1 - the profile of the sintering rate ζ at time tn+1 (see (6.8)).
The situation is unfortunately more complicated for the enthalpy and the temperature. Due to
the non-linearity in the enthalpy-to-temperature function β (6.2), standard implicit scheme
lead to non-linear equations to be solved at each time step and are therefore not appropriate.
The so-called Nochetto-Verdi (NV) scheme [83] combines the advantages of stability and
computing simplicity with only one linear equation to be solved at each time step. This is the
reason why our numerical scheme is based on this technique [38].
The NV scheme works in two steps. At first, it updates the temperature profile by solving a
Laplace problem for T n+1 which involves the enthalpy profile un only. This problem comes
from (6.13) and reads:
ρT n+1 −ωNVτdiv(k∇T n+1) = ρβ(un)+ωNVτF n+1, inΩ,
−k∂T n+1
∂n= κ(T −T0)+ f n+1, on ∂Ω,
(6.13)
with F n+1, f n+1 :Ω→R - the profiles of the sources F and f at time tn+1 and ωNV a relaxation
parameter. Stability is ensured [83] if ωNV is chosen bigger than the derivative of the function
β (see eq. (6.3)):
ωNV ≥ maxdβ
du= 1
minCp. (6.14)
The next step is to recover un+1 from un and from the temperature profile T n+1 computed in
the first step. The equation we used is based on (6.2) and reads:
un+1 = un + T n+1 −β(un)
ωNV, inΩ. (6.15)
56
6.2. Numerical method
hΦx
hΦy
hΦz
Figure 6.2 – Illustration of the net of parallelepipedic cells for the space discretization of thesintering potentialΦ.
The NV scheme however suffers form a serious drawback. The temperature profiles T n (called
Nochetto-Verdi temperature profiles) show only weak convergence to the real temperature
profiles. Fortunately, the convergence property of the enthalpy un is much better. For a
reliable approximation of the temperature profile at time tn it is therefore recommended to go
back to equation (6.2) and to use
ϑn =β(un) (6.16)
instead of T n .
6.2.2 Space discretization
A finite element method (FEM) is used to discretize the sintering potential, the enthalpy and
the temperature profiles.
The case of the sintering potential is again the simplest. Since equation (6.12) does not
involve any spatial derivation, it is possible to define the finite element approximation Φnh
of the sintering potential profile Φn as a piecewise constant function over a mesh of the
domainΩ. The mesh we propose forΦ is illustrated on Fig. 6.2 and consists of (possibly non
homogeneous) parallelepipedic cells.
57
Chapter 6. Numerical Modelling of SLS/SLM
The shape of the fabricated part is strongly connected to the sintering potential. Going back to
the definition ofΦ (6.4), we can identify the fabricated part with the subdomain ofΩ in which
Φ' 1. The mesh size hΦx ,hΦy and hΦz used to construct the discretization grid for Φ (Fig. 6.2)
turn out to be directly connected to the geometrical resolution of our numerical model. To
obtain predictions useful for general applications in mechanics, an accuracy of at least one
hundredth of a mm is required meaning that
hΦx = hΦy = hΦz ' 10µm. (6.17)
A consequence of this condition is the large number nΦ of degrees of freedom which are
needed forΦ at the scale of the domainΩ (some cm in the three directions). However, since
the sintering potential can be updated in each FEM cell separately (eq. (6.12)), the algorithmic
complexity only grows linearly with the number nΦ. A very fine grid (up to 109 cells) forΦ is
therefore possible with a limited impact on the method efficiency.
The situation is completely different for the discretization of the enthalpy and temperature
profiles. Because of the Laplace problem (6.13), we have to limit ourselves to a few hundred
thousand degrees of freedom to represent unh or T n
h if we want to maintain the algorithmic
complexity in an acceptable range. The finite element mesh to be used for unh or T n
h has to be
much coarser than the net (see Fig. 6.2) forΦnh .
To satisfy this condition and to be still sensitive enough to the fine details of the laser beam
shape (the typical beam radius are of some tens of µm), we propose a non-conformal mesh
evolving in time (Fig. 6.3). It is composed of a refined part following the laser completed
by a coarse part discretizing the entire domain Ω. The elements in the coarse mesh are
tetrahedra, the elements in the fine mesh are orthogonal prisms obtained by extruding the
surface triangles (see Fig. 6.3) in the z direction.
The dimensions H Tx , H T
y of the elements in the coarse mesh in the plane (x, y) (Fig. 6.3) are
mostly defined by the scanning strategy and by the laser beam radius ω. For focused laser
beams with ω< 100 µm we propose:
H Tx = H T
y ' ω
2. (6.18)
A stability condition is actually linking the choice (6.18) to the time step τ (6.10). It reads:
τ¿maxH T
x , H Ty
υ(6.19)
where v is the laser scanning speed. The idea of this condition is to prevent the laser from
flying over many coarse cells during a single time step. For normal processing condition (see
Tabs. 3.4 and 3.5), the stability condition (6.19) is easily fulfilled by the time step given in (6.10)
and the mesh sizes (6.18).
58
6.2. Numerical method
The parallelepipedic grid for the sintering potential (Fig. 6.2) is finally adapted to the coarse
mesh. We recommend to create it by dividing each coarse mesh intervals in 3 to 5 sub-intervals.
The size of the elements in the fine mesh is determined according to the heat source distri-
bution. Usually the regions with high temperature gradients are localized in the proximity of
the heat source. However, due to the significant difference between powder and bulk thermal
conductivities, the fine mesh should cover not only the laser spot area, but also the region
around it. We propose the following relation between the dimensions D of the fine mesh
(Fig. 6.3) and laser spot radius ω:
D
ω' 10−20. (6.20)
In the center of the fine mesh, where the laser is supposed to be located, we propose to select
the area A of the triangular element on the surface (Fig. 6.3) in order to have a good description
of the laser beam:
πω2
A' 20−30. (6.21)
The size of the elements in the fine mesh can be increased from the center to the border. The
idea is to smoothen the transition to the coarse mesh and to avoid too much hanging nodes
[84]. A thumb rule is that the ratio between triangle areas in neighboring zones should not
exceed 4.
In the z direction the discretization is made in a way that all the nodes are located on horizontal
planes (called reticular planes) with altitudes (Fig. 6.1)
z0 = 0 < z1 < z2 < z3 . . . < zm = hsubs. +hpowd. (6.22)
possibly spaced in a non-regular manner. To increase computation accuracy, the altitudes
zi are usually arranged in geometric progression. For the case of a thin powder layer on a
massive substrate we recommend to increase the z-discretization near the upper surface
where we have the laser activity. In this region, the fine mesh interval hz should be adapted to
the penetration depth (see § 3.3.1 of Chapter 3). We propose:
hz ' δ
10. (6.23)
In the vicinity of the interface between powder and substrate (where we expect properties
discontinuities), the mesh should also be refined in the z−direction. In order to prevent
numerical instabilities, the lengths of the last interval in the powder layer and the first interval
in the substrate should be of the same order. Except the upper elements, discretization of the
substrate is less important than in the powder domain.
59
Chapter 6. Numerical Modelling of SLS/SLM
H Ty
H Tx
D
Fine mesh following the laser, mesh size hT
Laser beam
Figure 6.3 – Top view of the non conformal mesh for the temperature/enthalpy profiles.
In conformity with the NV scheme (6.13) the enthalpy profile unh will be searched in the space
of piecewise constant functions over the mesh. The degrees of freedom for the enthalpy are
in the center of the elements. The temperature profile has to be more regular. We will look
for it under the form of a continuous piecewise linear1 function. The degrees of freedom for
the temperature are located at the nodes of the mesh. As it is usual for non-conformal FEM,
temperature discontinuities are tolerated across the border between the coarse and the fine
meshes. The continuity is imposed afterwards by a penalty technique [84].
1In the fine mesh where prismatic elements with 6 nodes are used, the temperature will be piecewise linearbut separately in the horizontal variables x, y and in the vertical variable z to match the number of degrees offreedom.
60
7 Sintering Threshold in SLS/SLM
The powder sintering threshold is one of the key parameters, required for the estimation of the
material state during the process (see Chapter 6). In particular, the knowledge of the sintering
threshold allows to predict dimensions of the final part. Nowadays the most-used criterion of
the sintering quality is based on the material melting temperature. However the discrepancies
between theoretical and experimental results for actual process parameters (i.e. high laser
powers and high scanning speeds, see § 2.1.3 of Chapter 2) point out that this criterion should
be improved.
In this Chapter we propose and justify a new method for the estimation of the powder sintering
threshold. This approach is based on the study of the melting of a single powder grain, which
has been presented in Chapter 4.
61
Chapter 7. Sintering Threshold in SLS/SLM
Average temperature, C
Pro
po
rtio
no
fliq
uid
0.0 500 1000 1500 2000 2500 30000.0
0.2
0.4
0.6
0.8
1.0
100 W
200 W
300 W
T SGMt
Figure 7.1 – Evolution of the molten material fraction and the average temperature of a18Ni(300) Maraging Steel powder grain for different laser powers.
7.1 Evolution of the grain melting threshold
In Chapter 4 we studied the melting process of a separate 18Ni(300) Maraging Steel powder
grain, irradiated by a plane wave (see Remark 7.1). The wave intensities corresponded to the
intensities of the laser beam with the properties listed in Tab. 3.4. We have chosen a range
of laser powers between 50 and 400 W, representative for actual industrial SLS/SLM process
parameters.
Remark 7.1. The values of the plane wave intensity I0 are always calculated for a laser beam of
radius ω= 40 µm: I0 = P/π/ω2 (see eq. (4.4) in Chapter 4). The laser wavelength is 1070 nm.
Laser power P Wave intensity I0 (4.4) T SGMt
50 W 9947 W/mm2 1021 C
100 W 19894 W/mm2 710 C
200 W 39789 W/mm2 442 C
300 W 59683 W/mm2 309 C
400 W 79577 W/mm2 269 C
Table 7.1 – Melting threshold T SGMt of a 18Ni(300) Maraging Steel powder grain, irradiated by
a plane wave of intensity I0, approaching the laser of power P .
62
7.1. Evolution of the grain melting threshold
Wave intensity I , W/mm2
Mel
tin
gth
resh
old
TSG
Mt
,C
0.0 10000 20000 30000 40000 50000 60000 70000 800000.0
300
600
900
1200
1500
approx.SGM
T∞t
T f
Is
T f −T∞t
e
Figure 7.2 – The dependency of the grain melting threshold T SGMt on the wave intensity.
Under these conditions, we have shown that the melting of a single powder grain is not
homogeneous over the grain volume (see Fig. 4.4). In order to connect the evolution of single
grain properties to the modification of the parameters of the whole powder bed, we estimated
the average temperature of a grain during its melting. Fig. 7.1 shows the relation between
the molten material fraction and the average temperature in the particle. It proves that the
melting of a separate powder particle starts when its average temperature is significantly less
than the melting point of the material (T f = 1413C, see Tab. 3.2). We will denote this grain
melting threshold as T SGMt (‘threshold’). The values of T SGM
t for different wave intensities are
listed in Tab. 7.1. This information is also represented graphically in Fig. 7.2.
From this figure, it can be seen that the threshold temperature T SGMt decreases from the
melting temperature T f to an asymptotic value T ∞t . We assume an exponential decay and we
propose to fit the curve in Fig. 7.2 with the empirical formula
T SGMt (I ) = T ∞
t + (T f −T ∞t )e−I /Is , (7.1)
where Is denotes the wave intensity, for which T SGMt −T ∞
t decreases by facor e times (Fig. 7.2).
The fit parameters we get in this situation for a typical 18Ni(300) Maraging Steel powder grain
(see Tab. 3.6) are:
Is ' 22984 W/mm2,
T ∞t ' 226C.
(7.2)
63
Chapter 7. Sintering Threshold in SLS/SLM
Depth, µm
Gra
inm
elti
ng
thre
sho
ld,
C
0.0 20 40 60 80 1000.0
300
600
900
1200
1500
T f
100 W
200 W
300 W
Figure 7.3 – The dependency of the melting threshold on the depth for wave intensities,approaching laser powers 100 W, 200 W, 300 W.
As we already discussed in § 3.3.1 of Chapter 3, the propagation in z-direction of a gaussian
laser beam with properties listed in Tab. 3.4, in a powder can be described by a generalized
Beer-Lambert law up to a depth not exceeding some millimeters:
I (x, y, z, t ) ' Il (x, y, t )e−z/δ. (7.3)
In (7.3) Il (x, y, t ) is the incident laser beam intensity at location (x, y) and time t on the surface
of the material and δ is the optical penetration depth of the material.
Combining (7.1) and (7.3), we get a law expressing the threshold temperature as a function of
the position in the powder bed:
T SGMt = T SGM
t (x, y, z, t ). (7.4)
The dependency of the melting threshold in the depth for different laser powers is shown in
Fig. 7.3. The penetration depth is 20 µm (see § 3.3.1 of Chapter 3).
From Fig. 7.3 we can see that, for the chosen wave intensities and powder properties, the
melting threshold of a particle is significantly less than the melting point of the material T f
if the particle is located at a depth less than 100 µm from the surface of the powder layer. It
can also be observed that the threshold temperature increases up to the melting point of the
material T f because, in the depth, the grains are heated more smoothly and homogeneously.
64
7.2. Process parameters and material properties
0.4 mm
1.4 mm
Figure 7.4 – Exact laser path and explanation of the laser delay.
We now propose to use the value T SGMt for the threshold Tt involved in the evolution equation
for the sintering potential (6.7) – (6.8) or (6.9) in the limit case, when partial sintering can be
neglected. Observe that all the discussions in the rest of this Chapter will be based on the
limit case equation (6.9). The justification for that can be found in Chapters 4 and 5, where we
have shown that, for 18Ni(300) Maraging Steel powder (Tab. 3.6) processed under standard
SLS/SLM parameters (Tab. 3.4) the melting and interconnection time of particles is small
compare to the laser irradiation time.
Our idea is now to propose an experimental validation of the empirical expression (7.4) for
the sintering threshold. To do so, we will make use of this relation together with (7.1), (7.2)
and (7.3) to simulate the SLM construction of a thin wall of 400 µm thickness out of 18Ni(300)
Maraging Steel powder and with two different parameter sets. At the end, we will compare the
numerical prediction of the interlayer connection quality to an EBSD analysis of the real parts
and to equivalent simulation results obtained with the default choice Tt = T f for the sintering
threshold.
7.2 Process parameters and material properties
The process conditions are listed in Tab. 7.2. We will denote the two sets of process parameters
as ‘Set I’ and ‘Set II’. The scanning strategy, used in both cases, is presented in Fig. 7.4. The
scanning starts with an angle of 30 from the border. When reaching the other border the laser
follows a rounded path (see Fig. 7.4). During this time it is switched off. The time of the laser
switch-off is 500 µs. The hatching distance of the strategy is 50 µm. The layer thickness is
also 50 µm. The powder is deposited on the solid massive substrate, which is also made of
18Ni(300) Maraging Steel.
Properties of 18Ni(300) Maraging Steel powder and liquid materials are presented in Tab. 3.6.
As we have already discussed in Chapter 6 (see eqs. (6.6) and (6.4)), we include into the simu-
lation the evolution of the absorptivity (from 0.8 to 0.2) and the effective thermal conductivity
(from 0.1 W/m/C to 26 W/m/C) of the material due to its phase change (see Chapter 8 for
more information about influence of these phenomena on the process).
65
Chapter 7. Sintering Threshold in SLS/SLM
Parameter Set I Set II
Laser wavelength 1070 nm 1070 nm
Laser power P 200 W 330 W
Laser beam radius ω 40 µm 40 µm
Laser beam profile TEM00 TEM00
Scanning speed υ 1.4 m/s 3.4 m/s
Hatching distance hh 50 µm 50 µm
Scanning angle 30 30
Powder layer thickness hl 50 µm 50 µm
Laser switch-off time 500 µs 500 µs
Table 7.2 – Sets of process parameters.
We also take the latent heat of fusion and the latent heat of vaporization of the material into
account (see § 3.2 of Chapter 3). The laser beam penetration depth is 20 µm (see § 3.3.1 of
Chapter 3).
The chosen substrate thickness is 250 µm, which is several times bigger, than the powder
layer (50 µm, see Tab. 7.2). The properties of the substrate correspond to the bulk 18Ni(300)
Maraging Steel (Tab. 3.6).
7.3 Simulation results
A single layer scanning has been modelled. The quality of the final part depends on the
attachment of the molten powder layer to the substrate or to the previous molten layer. In
order to estimate it, we compute the time spent by the material at a temperature higher than
the local melting threshold.
Fig. 7.5 represents the time spent above the local threshold as a function of the depth z, for
coordinates x and y corresponding to the middle of the scanning strategy. For comparison
the two parameters sets are analysed for the two different choices of the sintering threshold:
Tt = T SGMt (see (7.1), (7.2) and (7.3)) and Tt = T f .
Remark 7.2. The oscillations, seen on the graphs, are caused by the domain discretization. They
can be diminished by choosing of a finer mesh (see § 6.2.2 of Chapter 6 for the details). However
it will increase the simulation time and require high computer powers.
As it can be seen from Fig. 7.5, the default choice Tt = T f for the sintering threshold leads to the
conclusion that there is no attachment between the sintered region and the substrate for the
two sets of parameters (Fig. 7.5, dashed lines), which does not correspond to the experimental
results.
66
7.3. Simulation results
Depth, µm
Tim
eab
ove
thre
sho
ld,m
s
substrate
pow
der
layersu
rface
50 45 40 35 30 25 20 15 10 5 0.00.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Parameters I
Parameters II
Figure 7.5 – Time above the sintering threshold when Tt = T SGMt (solid lines) and when Tt = T f
(dashed lines) for Sets I and II (Tab. 7.2).
The time, spent by the material at a temperature above the sintering threshold, gives also an
indication about the quality of the final part. Fig. 7.5 shows, that the temperature near the
substrate stays above the threshold for a longer time (about 200 µs), when the parameters of
Set I are used. The conclusion is that the parameters of Set I are favorable to a better interlayer
attachment of the part and they should be preferred to the parameters of Set II.
The simulation results are in agreement with the study of the real sintered parts. Fig. 7.6 shows
the results of the EBSD (electron backscattered diffraction) analysis made on the polished
sections of the sintered parts in z-direction1.
The analysis represents the microstructure of the samples. It was made to see how the grains
have grown between different layers. Each color corresponds to a different crystallographic
orientation. The black regions are the regions which have not been indexed. That is caused by
the distortion of martensite crystals which do not always diffract the electrons clearly.
In Fig. 7.6, it can be seen that the grains of the sample ‘Set I’ are bigger than the grains in sample
‘Set II’. Moreover they are elongated in the direction of the construction. This experimental
observation confirms the prediction deduced from the simulation results presented in Fig. 7.5
on the quality of the interlayer attachment.
1The data are provided by Dr. J.Jhabvala.
67
Chapter 7. Sintering Threshold in SLS/SLM
Figure 7.6 – EBSD analysis of the results for sets I and II (Tab. 7.2). Provided by Dr. J.Jhabvala.
7.4 Conclusions
In this Chapter we discussed the questions related to the numerical estimation of the material
state during SLS/SLM. We have proposed a theory based on the results of the Single Grain
Model (Chapter 4), telling that the melting threshold of a grain in a scanned powder bed
depends on the local value of the incoming radiation intensity.
We justified this model by experimental validations. They are based on the comparison
between the observations by SEM/EBSD and the numerical predictions of the dimension and
of the interlayer connection in produced SLM samples.
The approach is not universal for all powders types and process parameters. Actually it requires
a preliminary analysis of the melting mechanisms of a single powder particle. However in
this Chapter we introduce an empirical formula (eqs. (7.1) – (7.2)) to estimate the sintering
threshold for 50 µm layer of the 18Ni(300) Maraging Steel powder (or other steel powders with
similar properties), sintered by a focused laser beam (Tab. 7.2) in the range of powers between
50 and 400 W and scanning speeds 1−3 m/s.
68
8 Mechanisms Affecting the Efficiencyof SLS/SLM Process
Additive processes have now reached a point where they are considered as possible alternatives
to traditional manufacturing technologies. At this moment their effectiveness becomes crucial.
That is why process efficiency is one of the key characteristics for SLS/SLM.
Application of high laser powers (up to 1 kW) can improve the process efficiency. However
significant energy losses and fast heat evacuation from the surface of the material during
SLS/SLM could reduce the benefit that the increase of the laser power should have on the speed
of the process. It could also cause premature wear of the technical equipment mounted in
immediate proximity of the powder surface: for example, back reflection of the laser irradiation
can damage lenses, scanning head, etc.
In this Chapter we discuss phenomena, which can be caused by the evolution of material
properties during SLS/SLM. We also study different mechanisms of energy losses and their
influence on the process efficiency.
69
Chapter 8. Mechanisms Affecting the Efficiency of SLS/SLM Process
P/υ Simulation Simulation Rosenthal (powder) Rosenthal (liq. mat.)
W/(m/s) αp = 0.8 α= 0.8−0.2 αp = 0.8 αl = 0.2
kp = 0.1 W/m/C k = 0.1−26 W/m/C kp = 0.1 W/m/C kl = 26 W/m/C
100/1 12700 C 2862 C 253895 C 3556 C
100/2 2862 C 2290 C 179559 C 2662 C
100/3 2862 C 1930 C 146617 C 2209 C
200/1 34200 C 2862 C 507790 C 7112 C
200/2 13200 C 2862 C 359118 C 5324 C
200/3 6250 C 2862 C 293234 C 4418 C
300/1 55900 C 3460 C 761685 C 10668 C
300/2 24200 C 2862 C 538677 C 7986 C
300/3 13700 C 2862 C 439851 C 6627 C
Table 8.1 – Influence of the thermal conductivity and absorptivity evolution on the maximumtemperature reached under the laser: simulation vs. Rosenthal solution (8.4).
8.1 Thermal conductivity and absorptivity evolution in SLS/SLM
In Chapter 3 we came to the conclusion that properties of a metallic powder can change
significantly during SLS/SLM. We have shown that in the case of 18Ni(300) Maraging Steel
powder, its absoprtivity evolves from quite a high value for the loose powder (0.7−0.8) to the
low one of the liquid metal (0.1−0.3). In the case of high laser powers it can cause significant
laser energy losses during sintering process and can even damage the equipment.
In Chapter 3 we also discussed that in the case of laser treatment of a bulk material the evolu-
tion of its thermal conductivity may be not significant, while in the case of metallic powders it
cannot be neglected. The difference between the thermal conductivities of 18Ni(300) Marag-
ing Steel powder and bulk material is more than 200 times: 0.1 W/m/C and 26 W/m/C for
powder and solid, respectively (see Tab. 3.6). We see that a metallic powder can act as an
insulator during the process. However its melting can provoke fast evacuation of laser heat
from the surface in depth of the material.
8.1.1 Simulation: constant material properties
To prove the existence of material powder absorptivity and thermal conductivity evolution
during SLS/SLM process, we will estimate numerically the maximum temperature reached
on the surface of a thick powder bed during laser scanning. Calculations are made in the
framework of the homogeneous medium hypothesis (see Chapter 6), in assumption that
melting does not influence the powder bed properties.
For the simulation we use the parameters corresponding to 18Ni(300) Maraging Steel powder
(Tab. 3.6). Thermal conductivity and absorptivity of the material stay constant: αp = 0.8,
kp = 0.1 W/m/C. Latent heat of fusion L f and latent heat of vaporization Lv are taken into
70
8.1. Thermal conductivity and absorptivity evolution in SLS/SLM
account (see § 3.2 of Chapter 3). Powder bed thickness is 250 µm, scanning strategy is a line.
As far as we deal with a powder, the laser is considered as a volumic heat source (the optical
penetration depth of powder is 20 µm, see Chapter 3). Laser beam radius is 40 µm, the profile
is top-hat.
Maximum temperature values reached on the surface of the powder bed during laser scanning
for the selected process parameters (see Tab. 3.5) are presented in the second column of
Tab. 8.1.
It is experimentally known that for the used process and material parameters the vaporization
point of a material is almost never reached (Tv = 2862 C, see Chapter 3). However the tem-
perature values, calculated in assumption of non-evolving material properties, significantly
exceed it.
Therefore during SLS/SLM, heat evacuation from the surface into the depth of a powder bed
due to a material thermal conductivity increase necessarily takes place. At the same time a
drastic decrease of the surface absorptivity of the material under the laser due to its melting
has also to be considered.
8.1.2 Rosenthal solution
We can estimate temperature under the laser by means of the Rosenthal solution of the heat
transfer equation for a semi-infinite material with a moving non-point surface heat source.
Assume a point heat source of the power P , which moves with the scanning speed υ along
x-axis on the surface of a semi-infinite homogeneous material with the effective thermal
conductivity k, the thermal diffusivity η= k/ρ/Cp and the absorptivity α.
In that case, in a referential moving with the laser, the pseudo-stationary solution of the heat
transfer equation is:
T (x, y, z) = P
2kπ
e−υ(r+x)
2η
r, (8.1)
where r =√
x2 + y2 + z2.
For a non-point heat source with intensity distribution Q(x, y), the expression of the tempera-
ture (8.1) generalizes as:
TQ (x, y, z) = α
2kπ
∫R2
Q(ξ,ζ)e−
υ2η (p
(x−ξ)2+(y−ζ)2+z2+(x−ξ))√(x −ξ)2 + (y −ζ)2 + z2
dξdζ. (8.2)
71
Chapter 8. Mechanisms Affecting the Efficiency of SLS/SLM Process
For a homogeneous cylindrical heat source of power P and radius ω (top-hat beam profile):
Q(ξ,ζ) =
P
πω2 , ξ2 +ζ2 <ω2
0, ξ2 +ζ2 ≥ω2,
(8.3)
(8.2) can be used to compute the value of the temperature under the laser:
TQ (0,0,0) = P
kπωH
(υω
2η
). (8.4)
The function H is an elliptic integral:
H(s) = 1
2π
∫ 2π
0
1−e−s(1−cosθ)
s(1−cosθ)dθ. (8.5)
For a gaussian heat source of power P and radius ω:
Q(ξ,ζ) = P
πω2 eξ2+ζ2
ω2 (8.6)
we get
TQ (0,0,0) = P
2kpπω
K
(υω
2η
), (8.7)
where function K is:
K (s) = 1
2π
∫ 2π
0e
s2
4 (1−cosθ)2(1−erf
( s
2(1−cosθ)
))dθ. (8.8)
Using equation (8.4) for the top-hat laser beam, we calculate the values of the maximum
temperature, reached on the surface of the material under the laser, for 18Ni(300) Maraging
Steel powder and liquid material (Tab. 3.6). Laser beam radius is 40 µm. The results are
presented in two last columns of Tab. 8.1. We can observe that the Rosenthal solution (8.4)
cannot be applied for the low-conductive materials (as powders) and high laser powers, as far
as the results are far from reality.
For liquid metal, the temperature values, provided by (8.4), are significantly less, but still
exceed the vaporization temperature of the material Tv = 2862C. It can be explained by the
fact that the penetration of laser radiation into the material is not taken into account. It
does not affect the results in the case of laser melting of bulk metals, but becomes crucial for
powders (see § 3.3.1 of Chapter 3).
72
8.1. Thermal conductivity and absorptivity evolution in SLS/SLM
Power/speed, W/(m/s)
Tem
per
atu
re,
C
0.0
8000
16000
24000
32000
40000
48000
56000
100/1 100/2 100/3 200/1 200/2 200/3 300/1 300/2 300/3
sim., kp ,αp
eq. (8.4), kl ,αl
sim., k,α evol.
Figure 8.1 – Influence of the thermal conductivity and absorptivity evolution on the maximumtemperature reached on the surface of the material during laser scanning (see Tab. 8.1).
8.1.3 Simulation: evolving material properties
When the melting temperature T f is reached, molten material is produced and the rear-
rangement and interconnection of particles take place. Therefore, in the framework of the
homogeneous medium hypothesis, the effective thermal conductivity of the medium evolves
from a low value kp , corresponding to loose powder, to a high value ks for solid material
(§ 3.3.3 of Chapter 3). The same mechanism explains the evolution of the surface absorptivity
of a powder bed during laser melting. It decreases from a high powder absorptivity αp to the
low value αl of a liquid material (§ 3.3.1 of Chapter 3)
The evolution laws of a material thermal conductivity and absorptivity are related to the
melting dynamics of a powder bed and will be discussed in Chapters 4 and 5. In our numerical
model, we neglect the stage of partially-sintered material, as it is explained in § 6.1.3 of
Chapter 6. We assume that, when the temperature reaches the threshold value Tt equal to
the melting temperature T f in this case, the thermal conductivity and the absorptivity of the
medium immediately evolve from kp to ks and from αp to αs , respectively (see Chapter 6).
We repeat the simulation, performed in § 8.1.1, but now we will include the evolution of
the material properties. As far as we model the homogeneous medium with the effective
properties of 18Ni(300) Maraging Steel powder (Tab. 3.6), we consider that its absorptivity
decreases from 0.8 to 0.2, while the thermal conductivity increases from 0.1 to 26 W/m/C,
when the melting temperature T f = 1413C is reached in conformity with the assumption that
the partial sintering phase is negligible (see Chapter 6). Latent heat of fusion L f and latent
heat of vaporization Lv are also taken into account. Laser spot radius is 40 µm, beam profile is
top-hat, optical penetration depth is 20 µm, scanning strategy is a line.
73
Chapter 8. Mechanisms Affecting the Efficiency of SLS/SLM Process
Time, µs
Ab
sorb
edla
ser
pow
er,W
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
P = 300 W
P = 200 W
P = 100 W
0.0
30
60
90
120
150
180
210
240
270
300
t tr3 t tr
2 t tr1
υ= 1 m/s
υ= 2 m/s
υ= 3 m/s
Figure 8.2 – Absorption of laser power in the beginning of the scanning for 18Ni(300) MaragingSteel powder.
Results of this simulation are presented in the third column of Tab. 8.1. The obtained values of
the maximum temperature on the surface of the powder bed never exceed the vaporization
point of the material Tv = 2862C. It corresponds to experimental results for the selected
process parameters. Thus, for a precise modelling of SLS/SLM of a metallic powder, it is
necessary to take into account the evolution of its properties due to the phase change.
Fig. 8.1 visualizes the comparison between the data, obtained numerically, and the results,
provided by Rosenthal solution (8.4) for the liquid material properties in § 8.1.2 (see Tab. 8.1).
We can see that only the simulation, which takes into account the evolution of material
properties due to the phase change, gives reasonable results. However temperature values,
calculated by means of Rosenthal solution (8.4) for the liquid metal properties, are not far
from the numerical results. It allows to conclude that the evolution of the material properties
due to the phase change is fast under considered process conditions (Tabs. 3.6 and 3.5).
8.1.4 Influence of absorptivity evolution on SLS/SLM efficiency
For SLM, process efficiency is defined as the ratio between the heat source integrated over the
whole powder bed (absorbed power) and the laser power itself.
We have shown in Chapter 3, that the absorptivity of 18Ni(300) Maraging Steel powder de-
creases significantly due to the phase change. Melting of a powder surface under the laser can
therefore cause a significant loss of the process efficiency.
74
8.1. Thermal conductivity and absorptivity evolution in SLS/SLM
Laser power P t tr
100 W 4.2 µs
200 W 2.4 µs
300 W 1.8 µs
Table 8.2 – Stabilization time of the surface state within the laser spot.
As far as the laser power and the scanning speed are kept constant and the powder is con-
sidered as a homogeneous medium, the surface state within the laser spot experiences a
transition from powder to molten material in some time after laser irradiation start. We will
denote this time as t tr (‘transition’). Using the simulation we can estimate t tr for the selected
process conditions.
Fig. 8.2 shows the laser power absorption in the beginning of the scanning. The treated cases
again correspond to the process parameters used in industrial applications (Tab. 3.5). The
scanning strategy is a line, ω = 40 µm and the laser beam profile is top-hat. The material
properties correspond to 18Ni(300) Maraging Steel powder bed (Tab. 3.6) and the absorptivity
evolves from 0.8 to 0.2.
Fig. 8.2 also shows that, the values of the absorbed power are close to αl P , when the surface
state within the laser spot is completely stabilized. Therefore almost all the powder bed
surface under the laser within the laser spot is molten after time t tr. The measured values of
the transition time t tr are presented in Tab. 8.2. For the selected process parameters, t tr does
not exceed 5 µs.
Our conclusion is that, for 18Ni(300) Maraging Steel and the chosen process conditions and
material parameters (Tabs. 3.6 and 3.5), SLS/SLM process has quite a low efficiency.
In fact, the process efficiency decrease due to the surface state evolution is defined by the ratio
between powder and liquid material absorptivity. In the case of metallic powders, this ratio
goes up to 4−5.
75
Chapter 8. Mechanisms Affecting the Efficiency of SLS/SLM Process
8.2 Convection and radiation
In Section 8.1 we have already discussed energy losses which can take place during SLS/SLM
of a metallic powder due to its absorptivity evolution. However radiation and convection
phenomena also can influence the process.
On the top surface Σ of the powder bed (see Fig. 6.1), the boundary condition (eq. (6.1)) can
be written as:
k∂T
∂n= qrad +qconv. (8.9)
The notations qrad and qconv are used for the fluxes describing the energy losses due to
radiation and convection, respectively.
To estimate the effects of radiation and convection, we will compare qrad and qconv to the flux
qabs of electromagnetic energy coming from the laser (power P , radius ω) and absorbed by
the material:
qabs(x, y) =αl
Pπω2 , if (x, y) ∈Σl ,
0, otherwise,(8.10)
with Σl the laser spot (see Fig. 8.3). In eq. (8.10) αl denotes the absorptivity of the material in
liquid state: as it was shown in Section 8.1, the laser mostly interacts with molten material
which justifies the use of αl and not of αp (powder material) or of αs (solid bulk material).
The flux qconv corresponding to convective losses, can now be expressed (see eq. (6.1)) as:
qconv = κ(T −T0), (8.11)
where κ is the heat transfer coefficient from the heated material surface to the surrounding
atmosphere of temperature T0' 20 C
The flux qrad, corresponding to radiative losses, is given by Stefan-Boltzmann law:
qrad = ε(T )σ((T +273)4 − (T0 +273)4). (8.12)
In (8.12), ε(T ) is the emissivity of the material (actually, also dependent on its surface state),
and σ= 5.67 ·10−14 W/mm2/K4 – the Stefan-Boltzmann constant. Equation (8.12) assumes
that all the temperatures are given in C.
76
8.2. Convection and radiation
8.2.1 Convection
For the estimation of the influence of the convection from the surface of a powder bed during
SLS/SLM we need to know the heat transfer coefficient κ. We can calculate it, considering the
melt pool, formed during laser scanning, as a hot horizontal ‘plate’. This ‘plate’ interacts with
the atmosphere. There exist convection and conduction heat flows parallel to each other and
to the normal to the boundary surface of liquid metal and atmosphere. The ratio of convective
to conductive heat transfer is defined by Nusselt number Nul:
Nul =κLc
katm, (8.13)
where Lc – the characteristic length of the liquid metal ‘plate’ and katm – the thermal conduc-
tivity of the surrounding atmosphere.
According to [85], for the top surface of a hot liquid metal plate in a colder environment:
Nul = 0.54Ral1/4, 105 ≤ Ral ≤ 2 ·107, (8.14)
where Ral is the Rayleigh number.
We consider Ar as the protective atmosphere of SLS/SLM process: katm = 0.016 W/m/C.
Characteristic length of a plate representing a melt pool can be found as
Lc = Surf.area
Perimeter. (8.15)
To estimate Lc we need to know the geometry of the melt pool formed during laser scanning.
We can get an idea of the shape of a stabilized melt pool by drawing the isotherm TQ (x, y,0) =T f of the pseudo-stationary solution of the heat transfer equation for a moving non-point heat
source (8.2). The results are presented in Fig. 8.3. It shows the position of the melt pool with
respect to the laser spot for process parameters with maximum laser energy density: P = 300 W,
υ= 1 m/s. Material properties correspond to 18Ni(300) Maraging Steel bulk material (Tab. 3.6).
However Fig. 8.3 does not give the exact information about the melt pool surface area, as far
as the evolution of thermal conductivity and absorptivity of the material due to its melting are
not taken into account in (8.2). For more precise estimation we will use numerical modelling.
According to the simulation results, for process parameters P = 300 W, υ=1 m/s (ω= 40 µm,
top-hat laser beam profile) and for the material properties corresponding to the effective
properties of 18Ni(300) Maraging Steel powder (Tab. 3.6), the area of a stabilized melt pool is:
Amp ' 0.045 mm2. (8.16)
77
Chapter 8. Mechanisms Affecting the Efficiency of SLS/SLM Process
x, µm
y,µ
m
−250 −200 −150 −100 −50 0.0 50100
50
0.0
50
100
T f
T f < T < Tv
T = T0
laser spot Σl
melt pool Σmp
−→υ
Figure 8.3 – Approximate position of the stabilized melt pool with respect to the laser spot:P = 300 W, υ= 1 m/s, ω= 40 µm (by Mathematica software).
Considering the melt poolΣmp as an ellipse (Fig. 8.3) with the minor semi-axis b approximately
equal to the radius ω of the laser beam Σl (for the top-hat beam profile):
b 'ω. (8.17)
Then the major semi-axis can be found as:
a = Amp/πb. (8.18)
The characteristic length (8.15) of an elliptical melt pool is now expressed as:
Lc =Amp
π(3(a +b)−p(3a +b)(a +3b))
, (8.19)
where we use the Ramanujan approximation for the ellipse perimeter [86]. For our parameters
and with (8.16) – (8.18) we obtain:
Lc ' 31 µm. (8.20)
Using (8.13), (8.14) and (8.19), we obtain a range for the heat transfer coefficient κ:
κ= Nulkatm
Lc, 0.005 ≤ κ≤ 0.019
W
mm2C. (8.21)
78
8.2. Convection and radiation
We can now estimate the energy losses by convection during SLS/SLM. This will be done under
the rough assumption that the temperature drops rapidly in the low conductive loose powder
and also in the already sintered part. In that case, the only contribution to the convection
losses will come from the melt pool Σmp (see Fig. 8.3) and their relative effect can be measured
by the ratio:
ηconv ≡ÎΣmp
qconv(x, y)dxdyÎΣ qabs(x, y)dxdy
. (8.22)
The denominator in the right-hand side gives the absorbed laser powerαl P (8.10). To estimate
the numerator, we replace qconv by its value (8.11) and we observe that the temperature within
the melt pool does certainly not to exceed the vaporization point Tv . We obtain
ηconv ≤κ(Tv −T0)Amp
αl P. (8.23)
Taking the numerical values
Tv = 2862C, T0 ' 20C, (8.24)
(8.21) for κ, (8.16) for Amp into account and replacing αl and P by values used for Fig. 8.3:
αl = 0.2, P = 300 W, (8.25)
we come to the conclusion that
ηconv ≤ 1−4%. (8.26)
This estimation shows that the relative laser energy losses due to convection are negligible
as long as from the melt pool is the only contribution. The actual situation can be different
if the powder material and the already sintered part start to play a significative role in the
convection processes.
79
Chapter 8. Mechanisms Affecting the Efficiency of SLS/SLM Process
8.2.2 Radiation
As we can see from (8.9), the estimation of laser energy losses due to radiation phenomenon
requires the value of the material emissivity. An empirical formula for the emissivity of a
porous media (i.e. a powder bed) is proposed in [87]. According to this model, the emissivity
of a powder bed mainly depends on its porosity. For 18Ni(300) Maraging Steel powder (with
the porosity of about 50%) it is approximately 0.7−0.8. The emissivity of a powder can also
change due to its melting. The exact value of the emissivity of the liquid maraging steel is not
known, but the value for the emissivity of molten steel at temperature close to the melt point
can be used instead: εl = 0.28 [88]. However, since our goal is to prove that radiative losses
are negligible, we will only consider the ‘worst’ case, when the emissivity of the material is
maximum:
ε= 1. (8.27)
We will again assume a quick drop of temperature inside the powder and the sintered part, so
that only the melt pool contributes to radiative losses. Their effect can thus be estimated by
the ratio
ηrad ≡ÎΣmp
qrad(x, y)dxdyÎΣ qabs(x, y)dxdy
. (8.28)
The numerator gives again the absorbed laser power αl P (8.10). An upper bound for the
numerator can be found by going back to the expression (8.12) of qrad and by observing that
the melt pool is obviously below the material melting point:
ηrad ≤ σ((Tv +273)4 − (T0 +273)4)Amp
αl P. (8.29)
With the numerical values (8.24) for Tv and T0, (8.27) for ε and (8.16) for Amp, and taking the
values (8.25) for αl and P , we get
ηrad ≤ 0.5%. (8.30)
The conclusion is that the laser energy losses due to radiation are also negligible if the only
significative contribution is the melt pool.
80
8.3. Conclusions
8.3 Conclusions
In this Chapter we proved that in the case of SLS/SLM of metallic powders it is necessary to
take the evolution of absorptivity and thermal conductivity into account. Increase of material
thermal conductivity leads to efficient heat evacuation from the surface of the processed mate-
rial, while the decrease of the surface absorptivity during liquefaction significantly diminishes
the energy efficiency of the process. Using the example of 18Ni(300) Maraging Steel powder
(Tab. 3.6), we have shown that the absorptivity evolution reduces the process efficiency by a
factor of 4. The situation is similar for most of the metallic materials used in SLS/SLM.
Process productivity can be increased by the selection of operating parameters which promote
the efficient laser-powder coupling. Our simulation method is certainly an efficient tool for
identifying such process conditions.
We also estimated energy losses caused by radiation and convection from the powder bed
during processing. The conclusion is that these mechanisms can be neglected as long as
they are localized inside the melt pool: for 18Ni(300) Maraging Steel powder and chosen
process parameters (Tab. 3.5) they do not exceed 5% of the laser input. An open question
remains about the actual radiative and convective effects in the loose powder and in the
already sintered part.
81
9 Melt Pool Dynamics
In Selective Laser Melting the powder grains under the laser are completely molten and form a
liquid domain called melt pool. Evolution of the melt pool during the process, its interaction
with the substrate and the surrounding non-molten powder strongly affect the quality of the
final part.
Estimation of the melt pool geometry allows to adjust the process parameters and, therefore,
to increase its efficiency. Predictions of the melt pool surface area can be also useful for the
calibration of online monitoring systems for Selective Laser Melting.
In this Chapter we discuss the dynamics of the melt pool, formed in SLM of 18Ni(300) Marag-
ing Steel powder. In the framework of the homogeneous medium hypothesis we study the
stabilization of the melt pool and the evolution of its geometrical aspects during laser scanning.
We also compare laser melting of a loose powder to a bulk material and propose a theoretical
method, based on the transient solution of the heat transfer equation, to estimate the melt
pool stabilization time and distance.
83
Chapter 9. Melt Pool Dynamics
Power/speed, W/(m/s) t bmp, ms t pb
mp, ms t plmp, ms t th
mp, ms (ε= 1%) t bcmp, ms
100/1 0.18 1.6 0.2 0.424 0.36
100/2 0.05 0.4 0.1 0.175 0.15
100/3 0.04 0.8 0.08 0.107 0.08
200/1 0.28 4.5 0.64 0.7 0.67
200/2 0.12 2.3 0.16 0.305 0.27
200/3 0.07 1.0 0.08 0.19 0.15
300/1 0.5 9.5 0.8 0.963 0.93
300/2 0.18 2.5 0.3 0.429 0.4
300/3 0.14 1.5 0.16 0.272 0.24
Table 9.1 – Time to stabilize the melt pool for different process and material parameters.
9.1 Bulk material
We can simulate the melt pool behavior during laser sintering of various materials by means of
our numerical model, based on the homogeneous medium hypothesis (see Chapter 6). In this
section, using 18Ni(300) Maraging Steel as an example, we will estimate the time to stabilize
the melt pool1 for a bulk highly-conductive material.
Properties of bulk 18Ni(300) Maraging Steel are listed in Tab. 3.6. In the simulation we take
into account melting and evaporation of the material and its absorptivity evolution due to
the phase change: from 0.5 to 0.2, according to the data from Chapter 3. The heat source is
treated numerically as a surface heat source. This assumption is justified by the fact that the
penetration depth of the laser beam is small in comparison with the material thickness (see
§ 3.3.1 of Chapter 3). For the simplicity we choose a single line as scanning strategy. The laser
beam profile is top-hat and its radius ω= 40 µm. The material thickness is 250 µm.
Values of the time needed to stabilize the melt pool for these conditions, are presented in the
second column of Tab. 9.1. We denote them as t bmp (‘bulk’). Also Fig. 9.1a demonstrates the
melt pool behavior during the scanning for the process parameters 100/1, 100/2 and 100/3
(laser power/scanning speed).
From Tab. 9.1 and Fig. 9.1a we see that after the material melting, it takes less than 500 µs
for the melt pool to stabilize. Stabilization time t bmp is decreasing with the increase of laser
scanning speed.
1Further in the text, if we speak about a melt pool stabilization time, calculated by means of the simulation, wemean ‘stabilization time of the surface area of a melt pool’.
84
9.1. Bulk material
Time, ms
Mel
tpo
ols
urf
ace
area
,mm
2
0.0 0.1 0.2 0.3 0.4 0.50.0
0.002
0.004
0.006
0.008
0.01
100/1
100/2
100/3
(a) 18Ni(300) Maraging Steel bulk material.
Time, ms
Mel
tpo
ols
urf
ace
area
,mm
2
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.80.0
0.02
0.04
0.06
0.08
0.1
0.12
100/1
100/2
100/3
(b) 18Ni(300) Maraging Steel powder.
Figure 9.1 – Evolution of the melt pool surface area during laser scanning of 18Ni(300) MaragingSteel powder and bulk material (Tab. 3.6).
85
Chapter 9. Melt Pool Dynamics
Power/speed, W/(m/s) hpbmp, µm hl , µm t pl
mp, ms
100/1 48 32 0.2
100/2 28 19 0.1
100/3 22 15 0.08
200/1 66 44 0.64
200/2 41 27 0.16
200/3 34 22 0.08
300/1 76 50 0.8
300/2 57 38 0.3
300/3 48 32 0.16
Table 9.2 – Melt pool depth reached during laser melting of 18Ni(300) Maraging Steel powder.
9.2 Powder bed
In Fig. 9.1b the melt pool behavior during laser scanning of 18Ni(300) Maraging Steel powder
for process parameters 100/1, 100/2, 100/3 is shown. Material properties are listed in Tab. 3.6.
In the simulation we consider absorptivity (from 0.8 to 0.2) and thermal conductivity (from
0.1 to 26 W/m/C) evolution due to the phase change of the material. The latent heat of fusion
L f and the latent heat of vaporization Lv (see Tab. 3.2) are also taken into account.
As far as we deal now with a powder bed, we consider the laser beam as a volumic heat source.
The penetration depth is δ= 20 µm. All other parameters stay the same as we used before:
the laser beam profile is top-hat, its radius is 40 µm, we choose a single track as a scanning
strategy, and the powder bed thickness is 250 µm.
In the third column of Tab. 9.1 values of the melt pool stabilization time for the case of
18Ni(300) Maraging Steel powder melting are presented. They are denoted as t pbmp (‘powder
bed’).
From Tab. 9.1 and Fig. 9.1b we can see that stabilization of the melt pool in the powder takes
much longer (up to a factor of 10) than in the bulk material. The surface area of the melt
pool, formed in the powder, is also bigger. We can explain it by the fact that the powder
thermal conductivity is very low compare to the thermal conductivity of the bulk material.
Non-molten powder, which surrounds the melt pool, plays the role of an insulator, while in
the highly-conductive bulk material the heat is quickly evacuated from the molten region.
In the second column of Tab. 9.2 we can see the depth hpbmp of the melt pool created during laser
sintering of 18Ni(300) Maraging Steel powder bed. As one can observe, for the chosen process
and material parameters, melt pool depth does not exceed 80 µm. The results correspond to
the current tendency to use powder layer thicknesses of several tens of microns for industrial
SLM of metallic powders.
86
9.3. Powder layer on a substrate
Time, ms
Mel
tpo
ols
urf
ace
area
,mm
2
0.0 0.1 0.2 0.3 0.4 0.5 0.60.0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
tatt
powder bed
32 µm powder + substrate
Figure 9.2 – Melt pool evolution during melting of 32 µm powder layer on a substrate:P = 100 W, υ= 1 m/s.
9.3 Powder layer on a substrate
In SLS/SLM process the laser scanning is actually applied to a layer of powder on a solid
substrate. The thickness of a layer is adjusted according to the properties of the processed
materials. The purpose is to melt a powder layer completely in depth and to attach it to the
substrate (or to the previous sintered layer). A good connection between powder layers and
the substrate provides a high quality of the final part.
Substrates are often made of the same (or similar) materials as sintered powders. Similarity of
the material properties between the substrate and the molten powder material allows to reach
better cohesion between them.
We modelled the case of SLM process of a thin powder layer on a bulk substrate. In Fig. 9.2
(black line) we can observe the melt pool behavior during the scanning of 32 µm layer of
18Ni(300) Maraging Steel powder on a substrate, made of the same material. We compare
the results to the melt pool evolution in the case of sintering of a thick powder bed (red line).
Laser beam is top-hat with a power of 100 W and a radius of 40 µm, the scanning speed 1 m/s.
For these process parameters the melt pool depth hpbmp in the case of powder bed processing is
about 48 µm (see Tab. 9.2). We choose the powder layer thickness hl = 32 µm: it is equal to 2/3
of 48 µm and should provide good attachment of the molten layer to the substrate.
In Fig. 9.2 we can see the moment when the melt pool touches the substrate: we denote it
as tatt (‘attachment’). When the melt pool reaches the highly-conductive substrate (or the
previous layer), the evacuation of heat from its surface into the depth becomes much more
87
Chapter 9. Melt Pool Dynamics
efficient. It provokes rapid cooling of the melt pool and, as a consequence, the decrease of
its surface area and its earlier stabilization. The melt pool formed in a thick powder bed and
surrounded by the low-conductive powder continues to grow.
We modelled the sintering of a powder layer on a substrate for all the sets of process parameters
(Tab. 3.5). Each time, to have an attachment with the substrate, we adjusted the layer thickness
hl according to the values of the melt pool depth hpbmp, obtained for the powder bed in § 9.2.
The chosen values of hl are presented in Tab. 9.2. They correspond to 2/3 of hpbmp for each set
of process parameters.
Remark 9.1. For process parameters corresponding to the low laser energy density (for example,
100/3), hl is of the order of the diameter of a typical powder grain (d0 = 8.87 µm for 18Ni(300)
Maraging Steel powder). Producing a layer of this thickness will be difficult. As a conclusion,
such process parameters may not be convenient for SLS/SLM of 18Ni(300) Maraging Steel.
Values of the melt pool stabilization time t plmp (‘powder layer’) are presented in Tab. 9.2. For
comparison with the other results we also added t plmp in Tab. 9.1 (fourth column).
From Tab. 9.1 we can see that the values of t plmp are in good correspondence with the melt
pool stabilization time for the bulk material t bmp. It can be explained by the fact that the used
powder layer thicknesses are low. After reaching an attachment between the melt pool and
the substrate, the heat transfer process is similar to the one taking place in a bulk material.
Melt pool behavior strongly depends on the thickness of powder layer, process conditions and
material properties. It is however interesting to observe that, considering the cases of melting
a bulk material and a thick powder bed, we obtain a lower and an upper bounds for the melt
pool stabilization time of a powder layer on a substrate.
9.4 Theoretical model
In Section 8.2 of Chapter 8 we discussed theoretical methods to estimate the surface area of
the melt pool formed during the selective laser melting of a given material. For this purpose
we used the pseudo-stationary solution of the heat transfer equation for a non-point heat
source of power P moving with speed υ along x-axis on the surface of a semi-infinite material
(see eq.(8.2)). However this solution only gives information about an already stabilized melt
pool. When the laser starts heating a medium at room temperature T0, the melt pool does not
reach its stable state immediately. Our goal is to estimate the time necessary to stabilize it
in the case of a semi-infinite material with thermal diffusivity η= k/ρ/Cp and absorptivity α
treated by a laser of power P , moving with speed υ. However, to simplify the calculations, we
come back to the case of a point heat source.
88
9.4. Theoretical model
Coordinate x, mm
Tem
per
atu
re,
C
−1.0 −0.9 −0.8 −0.7 −0.6 −0.5 −0.4 −0.3 −0.2 −0.1 0.00.0
1000
2000
3000
4000
5000
6000
7000
Figure 9.3 – Temperature evolution according to (9.1): 0.1, 0.3, 0.5, 0.7, 0.9 ms (from the rightto the left) from the laser irradiation start: P = 100 W, υ= 1 m/s.
In a referential moving with the laser, it can be seen that the temperature T (t , x) at a point
lying on the laser scan line at a distance x from the laser spot is:
T (t , x,0,0) = αP
2kπ
(e−υ(|x|+x)
2η
|x| −
− 1
2x
(e−υx
η(1−erf
(υt −x
2pηt
))− (1−erf
(υt +x
2pηt
))))+T0.
(9.1)
The first term in the brackets on the right-hand sight is the pseudo-stationary solution of the
Rosenthal equation, already discussed in Chapter 8 (see eq.(8.4) in § 8.1.2):
Tp.s.(x,0,0) = αP
2kπ|x|e−υ(|x|+x)
2η . (9.2)
The second term in the brackets in (9.1) is a transient solution used to match T (0, x,0,0) with
the initial temperature distribution:
T (0, x,0,0) = T0. (9.3)
89
Chapter 9. Melt Pool Dynamics
Fig. 9.3 shows an example of the evolution of the temperature profile T (9.1) in the negative
part of x-axis during laser scanning for the laser power 100 W and the scanning speed 1 m/s.
The moments of time are 0.1, 0.3, 0.5, 0.7, 0.9 ms from the laser irradiation start.
Material properties are listed in Tab. 9.3. They correspond to bulk 18Ni(300) Maraging Steel
(see Tab. 3.6), however the absorptivity of the material stays constant during laser processing.
From Fig. 9.3 we can observe that every temperature profile ‘joins’ quickly an upper bound,
which corresponds to the pseudo-stationary temperature profile (9.2).
The melt pool is obviously the area bounded by the isotherm corresponding to the melting
temperature T f of the material. As it was shown above, the extremum coordinate of this area
−rmp (in the negative part of the x-axis) will depend on time. We can calculate it, solving the
equation:
T (t ,−rmp(t ),0,0) = T f . (9.4)
When t →∞, rmp(t) tends to its maximum r∞mp, which is a solution of the equation for the
pseudo-stationary temperature profile:
Tp.s.(−r∞mp,0,0) = T f . (9.5)
Therefore from (9.2)
r∞mp = αP
2kπT f. (9.6)
From (9.1), (9.4) and (9.6) we can conclude that the relative difference between rmp(t ) and r∞mp
rmp(t )− r∞mp
r∞mp
(9.7)
will be less than an arbitrary quality ε> 0, if t is bigger than the value [89]:
t (ε) = t∞+ 2η
υ2
1+√√√√1+ υ2t∞
η ln 12ε
ln1
2ε, (9.8)
where
t∞ =r∞
mp
υ. (9.9)
In this context we define the melt pool stabilization time t thmp (‘theoretical’) as the value t (ε),
corresponding to a quite small relative error ε, e.g. ε= 1%.
90
9.4. Theoretical model
Absorptivity α Thermal conductivity k Heat capacity Cp Density ρ T f
0.5 26 W/m/C 0.5 J/g/C 0.008 g/mm3 1413 C
Table 9.3 – Material properties used for the estimation of the melt pool stabilization time.
In the fifth column of Tab. 9.1 we present the values of the melt pool stabilization time t thmp,
calculated using (9.6), (9.8) and (9.9).
Remark 9.2. Formula (9.8) can be validated by means of simulation. For this purpose we again
model laser melting of bulk 18Ni(300) Maraging Steel. The simulation parameters are similar
to the ones that we used in Section 9.1 of this Chapter (Tab. 9.3). However for the justification of
(9.8) we assume the material properties to be constant. We also do not take the latent heat of
fusion and the latent heat of evaporation into account.
Simulation results are presented in the last column of Tab. 9.1. We denote them as t bcmp (‘bulk
constant’). They are in a good correspondence with the values of the melt pool stabilization time
t thmp, provided by formula (9.8).
From Tab. 9.1 we can see that the numerical results for the bulk material with evolving proper-
ties t bmp are of the same order as the results t th
mp of (9.8). Good correspondence between t plmp
and t bmp also means that the theoretical method can definitely be used for the evaluation of
a melt pool stabilization time in the case of laser sintering of a thin low-conductive powder
layer on a massive highly-conductive substrate.
We can also observe that the melt pool stabilization time in the powder bed t pbmp is much bigger
than it is predicted by (9.8). This discrepancy can be explained by the significant difference
between the thermal conductivities of the melt pool and the surrounding powder and by the
evolution of the material properties during melting, which are not taken into account in (9.8).
Therefore we can conclude that, for typical industrial SLS/SLM parameters, the theoretical
method, presented above, is not suitable for all types of materials. We cannot use this approach
for metallic powders. However it can be applied for the estimation of the upper bound of the
melt pool stabilization time in the case of materials with constant or insignificantly evolving
properties.
Looking at the results, presented in Tab. 9.1, we can see that for all the types of materials the
melt pool stabilization time decreases with the increase of the laser scanning speed. It can be
explained by the fact that the growth of the laser speed reduces the depth of the formed melt
pool (see Tab. 9.2) and shallow melt pools tend to stabilize. If the stabilization time decreases
with laser speed, we can expect that their product representing the stabilization distance will
be a constant. This point is discussed in the next section.
91
Chapter 9. Melt Pool Dynamics
Power/speed, W/(m/s) d bmp, mm d pl
mp, mm d thmp, mm (ε= 1%)
100/1 0.18 0.2 0.424
100/2 0.1 0.2 0.35
100/3 0.12 0.24 0.321
200/1 0.28 0.64 0.7
200/2 0.24 0.32 0.61
200/3 0.21 0.24 0.57
300/1 0.5 0.8 0.963
300/2 0.36 0.6 0.858
300/3 0.42 0.48 0.816
Table 9.4 – Melt pool stabilization distance for different process and material parameters.
9.5 Stabilization distance
The distance, traveled by the laser before the melt pool stabilizes, is called stabilization distance
and denoted as dmp:
dmp = υtmp. (9.10)
The theoretical model (9.8) implies:
tmp ∼ 1
υ, (9.11)
for large values of υ. Therefore we might expect from (9.10) that dmp will only depend on laser
power and not so much on scanning speed.
To check this assumption, we compared the values of the stabilization distance d thmp (‘theoreti-
cal’), provided by (9.8) with simulation results for the cases of laser sintering/melting of bulk
material d bmp (Section 9.1) and of powder layers on a solid substrate d pl
mp (Section 9.3). The
values of the melt pool stabilization distance for the three cases are presented in Tab. 9.4.
Tab. 9.4 confirms that the stabilization distance does not depend much on scanning speed,
especially at low laser power. It also shows that, for the chosen process and material param-
eters (Tabs. 3.6 and 3.4) the melt pool stabilization distance is in the order of hundreds of
micrometers, both for the bulk material and for the powder layer on the substrate. These
results are in correspondence with the theoretical values, provided by (9.8).
92
9.6. Different scanning strategies
50 µm
400 µm
Figure 9.4 – Scanning strategy used for the numerical modelling in § 9.6.
9.6 Different scanning strategies
For the study, presented above, we always consider a single track as scanning strategy. We
choose it in order to obtain precise data on the melt pool stabilization time for the selected
process parameters. However, in real conditions, the melt pool behavior can be much more
complicated.
As we see from Tab. 9.1, in the case of laser melting of a 18Ni(300) Maraging Steel powder layer
on a massive substrate, made of the same material, the melt pool stabilization distance is of
the order of several hundreds micrometers for process parameters close to industrial ones.
Another factor influencing the melt pool behavior during laser sintering is the complexity of
the scanning strategy. To show it, we will study the scanning strategy presented in Fig. 9.4.
It consists of parallel lines of 400 µm length each, the hatching distance is 50 µm. After the
scanning of one line, the laser is switched off during 250 µs. Then it starts moving in the
opposite direction. Such a strategy is typically used to build real parts like thin walls.
We apply this scanning strategy to a 32 µm layer of 18Ni(300) Maraging Steel powder on a
substrate made in the same material (Tab. 3.6). The details of all the process parameters are
summarized in Tab. 9.5.
Laser power P 100 WScanning speed υ 1 m/sBeam profile top-hatBeam radius ω 40 µmLayer thickness hl 32 µmHatching distance hh 50 µmLaser switch-off 250 µs
Table 9.5 – Simulation parameters for the scanning strategy presented in Fig. 9.4.
93
Chapter 9. Melt Pool Dynamics
Time, ms
Mel
tpo
ols
urf
ace
area
,mm
2
0.0 0.4 0.8 1.2 1.6 2.0 2.4 2.8 3.2 3.60.0
0.005
0.01
0.015
0.02
0.025
Figure 9.5 – Melt pool evolution during scanning of the strategy Fig. 9.4: P = 100 W, υ= 1 m/s.
Fig. 9.5 presents the evolution of the melt pool area. One can observe that the melt pool totally
disappear between the scanning of two successive lines. It means that the laser switch-off
time (250 µs) is enough to reach complete resolidification of the molten regions.
The powder layer thickness is adjusted in a way to reach an attachment between the melt pool
and the substrate (see Tab. 9.2). The moments, when the molten region touches the substrate,
can be clearly identified on the graph in Fig. 9.5. They correspond to the quick drop of the
melt pool area due to an enforced heat diffusion. As soon as it has reached the substrate, the
melt pool is quickly stabilized. The oscillations correspond to the discretization of the domain,
used in numerical modelling (see § 6.2.2 of Chapter 6).
Actually, the melt pool area profile of the first scanned line in Fig. 9.4 is similar to the one we
have already seen in Fig. 9.2. As we have calculated, the melt pool stabilization distance d plmp is
about 200 µm for a power of 100 W and a scanning speed of 1 m/s (see Tab. 9.4). Therefore the
length of the line (400 µm) is enough to observe a stabilized melt pool.
From Fig. 9.5 we can see that the melt pool surface area during the scanning of the first line
is bigger that for the next ones. The explanation is as follows. The hatching distance hh of
the strategy is 50 µm, the beam radius is 40 µm. Therefore the scanned track overlaps the
previously sintered area (see Fig. 9.6), which possess the properties of the bulk material and is
highly-conductive. It causes more efficient heat evacuation of laser energy and decreases the
melt pool surface.
94
9.6. Different scanning strategies
Figure 9.6 – Overlapping of scanned tracks for the strategy Fig. 9.4 and top-hat laser beam ofradius 40 µm.
Partial remelting of previously sintered regions during SLS/SLM provides better quality of
the final part. In this case the melt pool surface area depends on the overlap of scanned
tracks. However, for homogeneous scanning strategies, we can observe some kind of melt
pool stabilization (see Fig. 9.4).
There is a slight decrease of the melt pool area during the scanning of each line. It can be
explained by the fact that, in the beginning, the sintered area under the laser is slightly less
than in the middle of a line (see Fig. 9.6).
95
Chapter 9. Melt Pool Dynamics
9.7 Conclusions
In this Chapter, we discussed the melt pool stabilization during SLM, using 18Ni(300) Maraging
Steel as an example of metallic powder (Tab.3.6). We compared the melting of a bulk material,
of a thick powder bed and of a thin powder layer on a substrate for typical industrial process
parameters (Tab. 3.5).
In the framework of the homogeneous medium model we concluded that the time necessary
for the complete melt pool stabilization during laser melting can vary from several hundreds
of microseconds for bulk materials to tens of milliseconds for thick powder beds. Such a wide
range is caused by the significant difference between thermal conductivities of powder and
bulk materials.
We have shown that in the case of laser melting of a powder layer on a massive substrate, the
stabilization time of a formed melt pool strongly depends on its attachment to the substrate,
i.e. on the chosen thickness of the layer. The layer thickness should be adjusted for each set of
process parameters. For this purpose, the value of the melt pool depth, obtained during SLM
of a thick powder bed can be used as an upper bound for the layer thickness. For example, for
18Ni(300) Maraging Steel and our process parameters, we see that the maximum melt pool
depth will never exceed 80 µm.
In the case of a good attachment of the molten region to the substrate (or to the previously
sintered layer) the melt pool stabilization time is close to the value for the bulk materials –
several hundreds of microseconds for our parameters.
We proposed an analytical formula (see eq.(9.8)) to predict the time for the melt pool stabiliza-
tion in the case of selective laser melting of a bulk material or a powder layer on a substrate.
Based on the theoretical and numerical results, we presented the melt pool stabilization dis-
tance as a better criterion for the estimation of the melt pool evolution, as far as it is not too
much dependent on the laser scanning speed. It was shown that for the chosen process and
material parameters, the melt pool stabilization distance never exceed 1 mm.
We have also demonstrated that in real conditions the behavior of the melt pool during SLM
depends not only on the laser and material parameters, but also on the complexity of the used
scanning strategy and on the dimensions of the final part.
96
10 Conclusions
This work is devoted to a theoretical study of Selective Laser Sintering and Selective Laser
Melting of 18Ni(300) Maraging Steel powder under process conditions corresponding to
operating parameters used in practical applications. For this purpose we use a finite element
simulation tool, built specially for SLS/SLM. The applied numerical model is based on the
homogeneous medium hypothesis. However, it is demonstrated that, for the comprehensive
understanding of the process, it is not sufficient to rely only on macroscopic models.
Numerical modelling of SLS/SLM also requires
• preliminary analysis of the loose powder properties;
• study of the phenomena, taking place at the powder grain level: melting and particle
interaction in the presence of liquid phase.
In this work, we have proposed theoretical approaches, allowing to estimate effective powder
properties. The results were validated experimentally. It was shown that during sintering/melt-
ing of a metallic powder, the thermal conductivity and the absorptivity of the material change
drastically due to the phase changes. In the domain of powder melting dynamics, we have
developed a model to simulate of the melting process of a separate particle located on the
surface layer of a powder bed (Single Grain Model). We have also introduced a theoretical
method, which allows to estimate the interaction of molten powder grains, approaching their
interaction by a coalescence mechanism.
It is shown that, for the chosen material properties and SLM operating parameters, the powder
grain melting time is negligibly small (several microseconds) compared to the laser irradiation
time. The same conclusion can be made about the merging time of two molten particles,
which is less than 1 µs. Therefore, the stage of partial powder sintering can certainly be
neglected in the SLM process under typical conditions.
97
Chapter 10. Conclusions
Another important result is that the powder grain melting is not homogeneous. Under specific
process conditions, powder sintering starts at average temperatures significantly less than
the melting point of the material. To allow for this mechanism at the macroscopic level, we
have proposed a completely new phenomenological evolution law for the powder sintering
threshold depending on the local laser beam intensity. The approach has been experimentally
validated.
The results, obtained by means of these sub-models, have been applied in the principal
numerical model, based on the homogeneous medium hypothesis. This approach allowed us
to simulate SLS/SLM process and to predict the final part quality with much better precision,
than the commonly used macroscopic numerical models do. By means of our method, we
have studied the SLS/SLM efficiency. It was shown that the evolution of the material properties
during SLS/SLM strongly influences the process. Increase of material thermal conductivity
leads to fast heat evacuation from the surface of the material, while the decrease of the surface
absorptivity during liquefaction significantly diminishes process energy efficiency (factor 4−5
for metallic powders). We also estimated energy losses caused by radiation and convection
from the powder bed during laser melting. The conclusion is that these mechanisms can be
neglected as long as they are localized inside the melt pool.
The dynamics of the melt pool, formed during SLM, was also investigated. The dependency
of the melt pool stabilization process on the used material was studied. We concluded that
the time necessary for the complete melt pool surface area stabilization during laser melting
can vary from several hundreds of microseconds for bulk materials to tens of milliseconds for
thick powder beds. In the case of laser melting of a thin powder layer on a massive substrate,
the melt pool stabilization time is close to the value for the bulk materials.
We have shown that in SLM the powder layer thickness should be chosen for each set of
operating parameters. Numerical estimation of the melt pool depth, obtained during SLM of a
thick powder bed, is a reliable and fast method for the layer thickness adjustment.
We proposed an analytical formula to predict the time for the melt pool stabilization in the
case of SLM of a bulk material or a powder layer on a substrate. Based on the theoretical and
numerical results, we came to the conclusion that the melt pool stabilization distance can be
also used as a criterion for the estimation of the melt pool evolution.
By means of the simulation, we have also demonstrated that in real conditions the behaviour
of the melt pool during SLM depends not only on the laser and material parameters, but also
on the complexity of the used scanning strategy and on the dimensions of the final part.
In general, we have shown that our simulation software is an efficient tool for identifying
optimal SLS/SLM process conditions in term of energy efficiency and for the prediction of the
dimensions and the quality of the final sintered part. However, for a better accuracy of these
predictions, the mechanisms of the cooling process of a sintered part should be studied and
implemented into the numerical model. This must be the subject of a future investigation.
98
Bibliography
[1] J. Jhabvala, Study of the consolidation process under macro-and microscopic thermal
effects in Selective Laser Sintering and Selective Laser Melting. PhD thesis, École Polytech-
nique Fédérale de Lausanne (EPFL), 2010.
[2] MATWEB, “http://www.matweb.com/,” 2015.
[3] A. Gusarov and E. Kovalev, “Model of thermal conductivity in powder beds,” Physical
Review B, vol. 80, no. 2, p. 024202, 2009.
[4] V. Postovalov, E. Romanov, V. Kondrat’ev, and V. Kononenko, “Theory of transport in liquid
metals: Calculation of dynamic viscosity,” High temperature, vol. 41, no. 6, pp. 762–770,
2003.
[5] S. Thoroddsen, K. Takehara, and T. Etoh, “The coalescence speed of a pendent and a
sessile drop,” Journal of Fluid Mechanics, vol. 527, pp. 85–114, 2005.
[6] J. Kruth, B. Van der Schueren, J. Bonse, and B. Morren, “Basic powder metallurgical
aspects in selective metal powder sintering,” CIRP Annals-Manufacturing Technology,
vol. 45, no. 1, pp. 183–186, 1996.
[7] J. Kruth, X. Wang, T. Laoui, and L. Froyen, “Lasers and materials in selective laser sinter-
ing,” Assembly Automation, vol. 23, no. 4, pp. 357–371, 2003.
[8] I. Yadroitsev, P. Bertrand, and I. Smurov, “Parametric analysis of the selective laser melting
process,” Applied surface science, vol. 253, no. 19, pp. 8064–8069, 2007.
[9] W. Meiners, K. Wissenbach, and R. Poprawe, “Direct selective laser sintering of steel
powder,” in Proceedings of the LANE, vol. 97, pp. 615–622, 1997.
[10] C. Over, W. Meiners, K. Wissenbach, M. Lindemann, and G. Hammann, “Selective laser
melting: a new approach for the direct manufacturing of metal parts and tools,” Proceed-
ings of the International Conferences on LANE, pp. 391–398, 2001.
[11] T. Wohlers, Wohlers report 2005: rapid prototyping, tooling & manufacturing state of the
industry: annual worldwide progress report. Wohlers associates, 2005.
99
Bibliography
[12] J. Kruth, P. Mercelis, J. Van Vaerenbergh, L. Froyen, and M. Rombouts, “Binding mecha-
nisms in selective laser sintering and selective laser melting,” Rapid prototyping journal,
vol. 11, no. 1, pp. 26–36, 2005.
[13] J. Kruth, B. Vandenbroucke, J. Vaerenbergh, and P. Mercelis, “Benchmarking of different
sls/slm processes as rapid manufacturing techniques,” 2005.
[14] J. Kruth, G. Levy, F. Klocke, and T. Childs, “Consolidation phenomena in laser and powder-
bed based layered manufacturing,” CIRP Annals-Manufacturing Technology, vol. 56, no. 2,
pp. 730–759, 2007.
[15] P. Karapatis, A sub-process approach of selective laser sintering. PhD thesis, École Poly-
technique Fédérale de Lausanne (EPFL), 2002.
[16] M. Rombouts, “Selective laser sintering/melting of iron-based powders,” Katholieke
Univesiteit Leuven, Leuven, 2006.
[17] N. Tolochko, M. Arshinov, A. Gusarov, V. Titov, T. Laoui, and L. Froyen, “Mechanisms of
selective laser sintering and heat transfer in ti powder,” Rapid prototyping journal, vol. 9,
no. 5, pp. 314–326, 2003.
[18] B. Van Der Schueren, “Basic contributions to the development of the selective metal
powder sintering process,” University of Leuven, Leuven, 1996.
[19] P. Aggarangsi and J. Beuth, “Localized preheating approaches for reducing residual stress
in additive manufacturing,” in Solid Freeform Fabrication Proceedings, pp. 709–720, 2006.
[20] P. Mercelis and J. Kruth, “Residual stresses in selective laser sintering and selective laser
melting,” Rapid Prototyping Journal, vol. 12, no. 5, pp. 254–265, 2006.
[21] S. Kumar, “Selective laser sintering: recent advances,” in 4th Pacific International Confer-
ence on Applications of Lasers and Optics, Wuhan, China, p. 8, 2010.
[22] V. Beal, P. Erasenthiran, N. Hopkinson, P. Dickens, and C. Ahrens, “The effect of scanning
strategy on laser fusion of functionally graded h13/cu materials,” The International
Journal of Advanced Manufacturing Technology, vol. 30, no. 9, pp. 844–852, 2006.
[23] V. Beal, P. Erasenthiran, C. Ahrens, and P. Dickens, “Evaluating the use of functionally
graded materials inserts produced by selective laser melting on the injection moulding
of plastics parts,” Proceedings of the Institution of Mechanical Engineers, Part B: Journal
of Engineering Manufacture, vol. 221, no. 6, pp. 945–954, 2007.
[24] S. Kumar, J. Kruth, J. Van Humbeeck, and A. Voet, “A study of degradation of laser-sintered
moulds using wear tests,” Rapid Prototyping Journal, vol. 15, no. 2, pp. 104–110, 2009.
[25] N. Raghunath and P. Pandey, “Improving accuracy through shrinkage modelling by using
taguchi method in selective laser sintering,” International journal of machine tools and
manufacture, vol. 47, no. 6, pp. 985–995, 2007.
100
Bibliography
[26] B. Caulfield, P. McHugh, and S. Lohfeld, “Dependence of mechanical properties of
polyamide components on build parameters in the sls process,” Journal of Materials
Processing Technology, vol. 182, no. 1, pp. 477–488, 2007.
[27] A. Pilipovic, T. Haramina, J. Kodvanj, and M. Šercer, “Influence of laser sintering parame-
ters on mechanical properties of polymer products,” in Proceedings of 3rd International
Conference on Additive Technologies, Hrvatska znanstvena bibliografija i MZOS-Svibor,
2010.
[28] M. Schmidt, D. Pohle, and T. Rechtenwald, “Selective laser sintering of peek,” CIRP
Annals-Manufacturing Technology, vol. 56, no. 1, pp. 205–208, 2007.
[29] A. Pilipovic, I. Drstvenšek, and M. Šercer, “Mathematical model for the selection of
processing parameters in selective laser sintering of polymer products,” Advances in
Mechanical Engineering, vol. 2014, 2014.
[30] L. Thijs, F. Verhaeghe, T. Craeghs, J. Van Humbeeck, and J. Kruth, “A study of the mi-
crostructural evolution during selective laser melting of ti–6al–4v,” Acta Materialia, vol. 58,
no. 9, pp. 3303–3312, 2010.
[31] S. Das, “Physical aspects of process control in selective laser sintering of metals,” Ad-
vanced Engineering Materials, vol. 5, no. 10, pp. 701–711, 2003.
[32] G. Cervera and G. Lombera, “Numerical prediction of temperature and density distri-
butions in selective laser sintering processes,” Rapid Prototyping Journal, vol. 5, no. 1,
pp. 21–26, 1999.
[33] A. Simchi, “Direct laser sintering of metal powders: Mechanism, kinetics and microstruc-
tural features,” Materials Science and Engineering: A, vol. 428, no. 1, pp. 148–158, 2006.
[34] B. Xiao and Y. Zhang, “Laser sintering of metal powders on top of sintered layers under
multiple-line laser scanning,” Journal of Physics D: Applied Physics, vol. 40, no. 21, p. 6725,
2007.
[35] T. Chen and Y. Zhang, “Three-dimensional modeling of selective laser sintering of two-
component metal powder layers,” Journal of Manufacturing Science and Engineering,
vol. 128, no. 1, pp. 299–306, 2006.
[36] L. JunChang, C. Langlade, and A. Vannes, “Evaluation of the thermal field developed
during pulsed laser treatments: semi analytical calculation,” Surface and Coatings Tech-
nology, vol. 115, no. 1, pp. 87–93, 1999.
[37] A. Nisar, M. Schmidt, M. Sheikh, and L. Li, “Three-dimensional transient finite element
analysis of the laser enamelling process and moving heat source and phase change
considerations,” Proceedings of the Institution of Mechanical Engineers, Part B: Journal of
Engineering Manufacture, vol. 217, no. 6, pp. 753–764, 2003.
101
Bibliography
[38] S. Kolossov, Non-linear model and finite element simulation of the selective laser sintering
process. PhD thesis, École Polytechnique Fédérale de Lausanne (EPFL), 2005.
[39] I. Roberts, C. Wang, R. Esterlein, M. Stanford, and D. Mynors, “A three-dimensional
finite element analysis of the temperature field during laser melting of metal powders in
additive layer manufacturing,” International Journal of Machine Tools and Manufacture,
vol. 49, no. 12, pp. 916–923, 2009.
[40] T. Childs and C. Hauser, “Raster scan selective laser melting of the surface layer of a tool
steel powder bed,” Proceedings of the Institution of Mechanical Engineers, Part B: Journal
of Engineering Manufacture, vol. 219, no. 4, pp. 379–384, 2005.
[41] A. Gusarov and J. Kruth, “Modelling of radiation transfer in metallic powders at laser
treatment,” International journal of heat and mass transfer, vol. 48, no. 16, pp. 3423–3434,
2005.
[42] A. Gusarov and I. Smurov, “Modeling the interaction of laser radiation with powder bed
at selective laser melting,” Physics Procedia, vol. 5, pp. 381–394, 2010.
[43] X. Wang, T. Laoui, J. Bonse, J. Kruth, B. Lauwers, and L. Froyen, “Direct selective laser
sintering of hard metal powders: experimental study and simulation,” The International
Journal of Advanced Manufacturing Technology, vol. 19, no. 5, pp. 351–357, 2002.
[44] C. Argento and D. Bouvard, “Modeling the effective thermal conductivity of random
packing of spheres through densification,” International journal of heat and mass transfer,
vol. 39, no. 7, pp. 1343–1350, 1996. provided by EPFL.
[45] A. Gusarov, T. Laoui, L. Froyen, and V. Titov, “Contact thermal conductivity of a powder
bed in selective laser sintering,” International Journal of Heat and Mass Transfer, vol. 46,
no. 6, pp. 1103–1109, 2003.
[46] J. Steinberger, J. Shen, K. Manetsberger, and J. Muellers, “The simulation of the sls
process as the basis of a process optimization,” in Proceedings of the 11th Solid Freeform
Fabrication Symposium, University of Texas, Austin, pp. 377–385, 2000. provided by EPFL.
[47] S. Safdar, L. Li, and M. Sheikh, “Numerical analysis of the effects of non-conventional
laser beam geometries during laser melting of metallic materials,” Journal of Physics D:
Applied Physics, vol. 40, no. 2, p. 593, 2007.
[48] A. Bauereiß, T. Scharowsky, and C. Körner, “Defect generation and propagation mech-
anism during additive manufacturing by selective beam melting,” Journal of Materials
Processing Technology, vol. 214, no. 11, pp. 2522–2528, 2014.
[49] Azom, “http://www.azom.com/,” 2015.
[50] A. Pinkerton and L. Li, “An analytical model of energy distribution in laser direct metal
deposition,” Proceedings of the Institution of Mechanical Engineers, Part B: Journal of
Engineering Manufacture, vol. 218, no. 4, pp. 363–374, 2004.
102
Bibliography
[51] EMPA, “http://www.empa.ch/,” 2015.
[52] S. Seshadri, “Effect of absorption on the spreading of a laser beam,” Optics letters, vol. 29,
no. 11, pp. 1179–1181, 2004.
[53] D. Bergström, “The absorption of laser light by rough metal surfaces,” 2008.
[54] R. Glardon, N. Karapatis, V. Romano, and G. Levy, “Influence of nd: Yag parameters on the
selective laser sintering of metallic powders,” CIRP Annals-Manufacturing Technology,
vol. 50, no. 1, pp. 133–136, 2001.
[55] J. Kruth, P. Peeters, T. Smolderen, J. Bonse, T. Laoui, and L. Froyen, “Comparison between
co2 and nd: Yag lasers for use with selective laser sintering of steel-copper powders,”
Revue internationale de CFAO et d’informatique graphique, vol. 13, no. 4-6, pp. 95–110,
1998.
[56] N. Tolochko, Y. Khlopkov, S. Mozzharov, M. Ignatiev, T. Laoui, and V. Titov, “Absorptance
of powder materials suitable for laser sintering,” Rapid Prototyping Journal, vol. 6, no. 3,
pp. 155–161, 2000.
[57] Y. Kizaki, H. Azuma, S. Yamazaki, H. Sugimoto, and S. Takagi, “Phenomenological studies
in laser cladding. part i. time-resolved measurements of the absorptivity of metal powder,”
Japanese journal of applied physics, vol. 32, no. part 1, pp. 205–212, 1993.
[58] S. Sih and J. Barlow, “The measurement of the thermal properties and absorptances of
powders near their melting temperatures,” in Proceedings of the Solid Freeform Fabrica-
tion Symposium 1992, pp. 131–140, DTIC Document, 1992.
[59] C. Van Gestel, Study of physical phenomena of selective laser melting towards increased
productivity. PhD thesis, École Polytechnique Fédérale de Lausanne (EPFL), 2015.
[60] M. Rombouts, L. Froyen, A. Gusarov, E. Bentefour, and C. Glorieux, “Light extinction in
metallic powder beds: Correlation with powder structure,” Journal of applied physics,
vol. 98, no. 1, p. 013533, 2005.
[61] M. Alkahari, T. Furumoto, T. Ueda, A. Hosokawa, R. Tanaka, A. Aziz, and M. Sanusi,
“Thermal conductivity of metal powder and consolidated material fabricated via selective
laser melting,” in Key Engineering Materials, vol. 523, pp. 244–249, Trans Tech Publ, 2012.
[62] Y. Shi and Y. Zhang, “Simulation of random packing of spherical particles with different
size distributions,” in ASME 2006 International Mechanical Engineering Congress and
Exposition, pp. 539–544, American Society of Mechanical Engineers, 2006.
[63] AKSTEEL, “http://www.aksteel.com/,” 2015.
[64] D. Bergström, J. Powell, and A. Kaplan, “The absorptance of steels to nd: Ylf and nd: Yag
laser light at room temperature,” Applied Surface Science, vol. 253, no. 11, pp. 5017–5028,
2007.
103
Bibliography
[65] J. Xie, A. Kar, J. Rothenflue, and W. Latham, “Temperature-dependent absorptivity and
cutting capability of co2, nd: Yag and chemical oxygen–iodine lasers,” Journal of Laser
Applications, vol. 9, no. 2, pp. 77–85, 1997.
[66] J. Xie and A. Kar, “Laser welding of thin sheet steel with surface oxidation,” WELDING
JOURNAL, NEW YORK, vol. 78, pp. 343–s, 1999.
[67] P. Fischer, N. Karapatis, V. Romano, R. Glardon, and H. Weber, “A model for the interaction
of near-infrared laser pulses with metal powders in selective laser sintering,” Applied
physics A, vol. 74, no. 4, pp. 467–474, 2002.
[68] G. Mie, “Beiträge zur optik trüber medien, speziell kolloidaler metallösungen,” Annalen
der physik, vol. 330, no. 3, pp. 377–445, 1908.
[69] A. Lozinskaïa and E. Boillat, “Potentiels de debye et diffraction par une sphère,” Internal
communication LGPP, 2009.
[70] M. Korolczuk-Hejnak, “Empiric formulas for dynamic viscosity of liquid steel based on
rheometric measurements,” High Temperature, vol. 52, no. 5, pp. 667–674, 2014.
[71] K. Laidler, “Chemical kinetics, 3rd,” 1987.
[72] N. Eustathopoulos, B. Drevet, and E. Ricci, “Temperature coefficient of surface tension
for pure liquid metals,” Journal of crystal growth, vol. 191, no. 1, pp. 268–274, 1998.
[73] Z. Li, K. Mukai, M. Zeze, and K. Mills, “Determination of the surface tension of liquid
stainless steel,” Journal of materials science, vol. 40, no. 9-10, pp. 2191–2195, 2005.
[74] W. Hager, “Wilfrid noel bond and the bond number,” Journal of Hydraulic Research,
vol. 50, no. 1, pp. 3–9, 2012.
[75] Y. Frenkel, “Viscous flow of crystalline bodies under the action of surface tension,” SCI.
SINTERING Sci. Sintering, vol. 12, no. 1, p. 7, 1980.
[76] J. Eggers, J. Lister, and H. Stone, “Coalescence of liquid drops,” Journal of Fluid Mechanics,
vol. 401, pp. 293–310, 1999.
[77] A. Menchaca-Rocha, A. Martínez-Dávalos, R. Nunez, S. Popinet, and S. Zaleski, “Coa-
lescence of liquid drops by surface tension,” Physical Review E, vol. 63, no. 4, p. 046309,
2001.
[78] Y. Shikhmurzaev, Capillary flows with forming interfaces. CRC Press, 2008.
[79] T. Iwashita, D. Nicholson, and T. Egami, “Elementary excitations and crossover phe-
nomenon in liquids,” Physical review letters, vol. 110, no. 20, p. 205504, 2013.
[80] P. Clausen, M. Wicke, J. Shewchuk, and J. O’Brien, “Simulating liquids and solid-liquid
interactions with lagrangian meshes,” ACM Transactions on Graphics (TOG), vol. 32, no. 2,
p. 17, 2013.
104
Bibliography
[81] S. Kolossov, E. Boillat, R. Glardon, P. Fischer, and M. Locher, “3d fe simulation for temper-
ature evolution in the selective laser sintering process,” International Journal of Machine
Tools and Manufacture, vol. 44, no. 2, pp. 117–123, 2004.
[82] T. Hughes, “Unconditionally stable algorithms for nonlinear heat conduction,” Computer
Methods in Applied Mechanics and Engineering, vol. 10, no. 2, pp. 135–139, 1977.
[83] E. Magenes, R. Nochetto, and C. Verdi, “Energy error estimates for a linear scheme
to approximate nonlinear parabolic problems,” RAIRO-Modélisation mathématique et
analyse numérique, vol. 21, no. 4, pp. 655–678, 1987.
[84] E. Boillat, “Finite element methods on non-conforming grids by penalizing the matching
constraint,” ESAIM: Mathematical Modelling and Numerical Analysis, vol. 37, no. 02,
pp. 357–372, 2003.
[85] M. Lindeburg et al., Mechanical engineering reference manual for the PE exam.
www.ppi2pass.com, 2013.
[86] S. Ramanujan, “Modular equations and approximations to π,” Quart. J. Math, vol. 45,
pp. 350–372, 1914.
[87] S. Sih and J. Barlow, “Emissivity of powder beds,” in Solid Freeform Fabrication Sympo-
sium Proceedings, vol. 6, pp. 7–9, Center for Materials Science and Engineering, Mechani-
cal Engineering Department and Chemical Engineering Department, the University of
Texas at Austin, 1995.
[88] T. of Emissivity, “http://www.monarchserver.com/tableofemissivity.pdf,” 2015.
[89] E. Boillat, “Une solution de l’équation de la chaleur sous la forme d’un profil translaté en
3d,” Internal communication LGPP, 2015.
105
CURRICULUM VITAE
PERSONAL INFORMATION
Name
Date of birth
TATIANA POLIVNIKOVA
27 APRIL 1986
Place of birth RUSSIAN FEDERATION
E-mail [email protected]
WORK EXPERIENCE
• Dates (from – to) January 2010 – July 2011
• Name and address of employer Electrosetstroyproekt, Moscow Russia
• Type of business or sector Research and development in energy transmission sector
• Occupation or position held researcher
EDUCATION
• Dates (from – to)
• Name and type of organization
• Dates (from – to)
July 2011 – September 2015
Laboratoire de Gestion et Procédés de Production (LGPP), Ecole
Polytechnique Fédérale de Lausanne
Doctoral student
September 2003 – February 2009
• Name and type of organization Moscow State University, Faculty of Physics, Physical Electronics
Department, Moscow, Russian Federation
• Title of qualification awarded Master Degree in Physics
107